ESMRMB

European Society for Magnetic Resonance in Medicine and Biology

Magnetic Resonance Materials in Physics, Biology and Medicine

magma

Suppl. 1, Vol. 16, 2003

Book of Abstracts ESMRMB 2003 20th Annual Scientific Meeting Rotterdam/NL, Sept. 18-21, 2003 ISSN electronic edition: 1352-8661 DOI: 10.1007/s10334-003-0017-5

Official Journal of the European Society for Magnetic Resonance in Medicine and Biology

S2

Basic MRS / Introduction to Functional MRI

ORAL PRESENTATIONS Thursday, September 18, 2003

Teaching Session I 8:30 am - 10:00 am

Willem Burger

Basic MRS 1 1H MR spectroscopy: indications and techniques A. D. Waldman; Imaging, Charing Cross Hospital, London, UNITED KINGDOM. Major clinical applications of proton MR spectroscopy in neuroscience will be reviewed. These will include non-invasive diagnosis of focal brain lesions, degenerative brain diseases, and inflammatory and metabolic disorders. Particular emphasis will be given to the ‘added value’ of MRS when used as an adjunct to structural imaging. The relative merits of widely available acquisition and localisation techniques in different clinical settings will be discussed; PRESS and STEAM, short and long echo-times, single voxel and spectroscopic imaging. Quantitative and semi-quantitative approaches to data analysis, and issues of discriminant power and reproducibility relevant to clinical applications will also be addressed.

2 1H MR spectroscopy at 3 T: improvements and limits U. Klose; Section of exp. NMR of the CNS, University Hospital, Tübingen, GERMANY. Introduction: In the past, clinical spectroscopic studies were mostly performed at a field strength of 1.5 T. In the recent years, a number of 3T whole body systems have been installed at clinical sites. The possible improvements and disadvantages of spectroscopic measurements of patients at 3 T are discussed. Methods: As at 1.5 T, the most common MR spectroscopy techniques in in-vivo examinations at 3 T are volume selective sequences (PRESS and STEAM) and chemical shift imaging techniques. The advantages and special difficulties of both approaches will be summarized and illustrated by published results and own experiments. Results: Beside the improved signal-to-noise ratio at 3 T, the advantage of higher field strengths in spectroscopy is the increased frequency range of chemical shifts of metabolic resonances. This is partly compromised by larger linewidths due to reduced T2* relaxation times which results in a reduced spectral resolution. Nevertheless, distinct improvements are obtainable especially in the identification of metabolites with coupled spinsystems like glutamine, glutamate and GABA. The frequency difference between the single lines of the complex multiplet structure remains unchanged at higher field strengths and the extent of superposition of signals of different metabolites is reduced. A disadvantage of 3 T is however that additional problems arise due to the larger inhomogeneity of the magnetic field in certain regions like the frontal lobe in the brain. Discussion: Spectroscopic measurements are one of those applica-

tions in clinical used magnetic resonance techniques, where the increase in field strength from 1.5 T to 3T has strong advantages, as the detectability of coupled substances are clearly improved. For standard examinations with long echo times, however, occurring problems with local field inhomogeneities and faster signal decay might diminish the advantages. A change to 3 T is therefore not necessarily indicated for all spectroscopic examinations, but it can improve the sensitivity to detect metabolites with coupled signals, which are difficult to examine at 1.5 T.

3 1H MR spectroscopy in brain trauma M. R. Garnett; Biochemical and Clinical Magnetic Resonance Unit, John Radcliffe Hospital, Oxford, UNITED KINGDOM. Purpose / Introduction: Traumatic brain injury (TBI) is a significant cause of morbidity and mortality. Clinical recovery following TBI is a protracted process and may be incomplete, even in patients who have no abnormalities on conventional (T1 or T2 weighted) imaging. Microscopic diffuse axonal injury, not readily detectable on conventional imaging, is a possible explanation for the lack of clinical recovery. 1H MR spectroscopy can be used to longitudinally study this cellular injury in patients following TBI. Subjects, Methods and Results: Data obtained by the Biochemical and Clinical Magnetic Resonance Unit, Oxford will be discussed. In summary, 1H MR spectroscopy was acquired from normal appearing frontal brain in a cohort of 27 patients. Clinical and 1H MR spectroscopy data was acquired on 3 occasions over a 2 year period. There was a significant reduction in Nacetyl aspartate (NAA) together with significant increases in choline compounds (Cho) and myo-inositol (Ins) during the 2 year period. In addition there was a significant correlation between the reduction in NAA and clinical recovery. Conclusions: 1H MR spectroscopy following TBI yields information that is not obtainable using conventional imaging. The 1H MR spectroscopy is reporting on microscopic cellular injury, the extent of which reflects the potential clinical recovery of the patient.

Teaching Session I 8:30 am - 10:00 am

Jurriaanse

Introduction to Functional MRI 4 Basics of BOLD fMRI R. Turner; Wellcome Department of Imaging Neuroscience, Institute of Neurology, London, UNITED KINGDOM. BOLD contrast fMRI relies on the relative paramagnetism of deoxyhaemoglobin, compared with all other tissues, to provide changes in MRI signal indicating localized brain activity. In order to use this phenomenon in brain science research, the relationship between MR image intensity changes and changes in neuronal activity must be understood. Appropriate means of acquiring MRI data and performing statistical analysis must also be adopted. The basis of BOLD has been explored using a number of techniques. These include fMRI, arterial spin labelled perfusion MRI, intracranial optical imaging, multielectrode recording of neuronal activity, and mathematical modelling, to enable a synthesis of

Cardiac MRI/MRS: a new frontier knowledge giving a good understanding of the underlying biophysics, which will be discussed. Echo-planar imaging or equivalent methods are used to obtain very rapid images of brain volumes free from motion artifact, with specific preferred imaging parameters. A detailed rationale for the choice of imaging parameters will be presented. Magnetic susceptibility variations within the head can cause serious image distortion and dropout. Methods for minimizing dropout while maintaining BOLD sensitivity will be described, and a practical algorithm for removal of image distortion and its interaction with head movement will be outlined. Image reconstruction methods minimizing image ghosting will be discussed, together with methods of task presentation and data analysis, and some comments on robust experimental design. Typical cognitive fMRI studies involving taste and smell stimuli, and memory will be described.

5 Basics and application of high field fMRI M. Barth; Univ.Klinik f. Radiodiagnostik, Universitaet und AKH Wien, Wien, AUSTRIA. During the last years MR scanners suitable for human examinations have been used for functional MRI of the brain at high magnetic field strengths, i.e. from 3 Tesla up to 7 Tesla. Besides the inherent increase in the signal-to-noise ratio (SNR) other advantages such as higher BOLD contrast-to-noise ratio (CNR) were expected and found in functional studies at these high field strengths. This enhanced SNR and CNR can be used to improve spatial and/or temporal resolution, to shorten measurement time or to improve statistical analysis [1,2]. However, high fields also have some drawbacks such as lower homogeneity of the main magnetic field (B0) and higher sensitivity towards physiological artifacts. In addition, different relaxation times and higher RF-power deposition have to be accounted for. Nevertheless, fMRI at field strengths of 3 and 4 Tesla have been found to result in a more stable and reproducible activation compared to 1.5 Tesla [3,4], which is very important for emerging clinical applications such as presurgical fMRI. Higher fields can be used for submillimeter resolution fMRI studies and seem to offer higher specificity towards the actual site of neuronal activation. Also non-BOLD-based acquisition methods, such as spin-echo based or proton-density weighted sequences [5,6], benefit from the higher sensitivity. References: 1. Thulborn KR et al. High-resolution echo-planar fMRI of human visual cortex at 3.0 tesla. NMRBiomed(1997) 2. Pfeuffer J et al. Perfusion-based high-resolution functional imaging in the human brain at 7 Tesla. MRM(2002) 3. Kruger G et al. Neuroimaging at 1.5 T and 3.0 T: comparison of oxygenation-sensitive magnetic resonance imaging. MRM(2001) 4. Yang Y et al. Comparison of 3D BOLD functional MRI with spiral acquisition at 1.5 and 4.0 T. Neuroimage(1999) 5. Yacoub E et al. Spin-echo fMRI in humans using high spatial resolutions and high magnetic fields. MRM(2003) 6. Stroman PW et al. Functional magnetic resonance imaging of the human brain based on signal enhancement by extravascular protons (SEEP fMRI). MRM(2003)

S3

6 fMRI of the visual system C. Segebarth, PhD; Unité Mixte INSERM / UJF 594, LRC CEA 30V, INSERM, Grenoble, FRANCE. Introduction: The visual system in humans is composed of numerous functionally distinct cortical areas which are localised in the occipital pole of the brain and which are to some extent organised in a hierarchical fashion. The “low order” visual areas, besides, are organised “retinotopically”, i.e. each region of the visual field is represented by one particular sub-area within each of these visual areas. Also, regions nearby within the visual field are represented by sub-areas which are equally positioned nearby on the cortical surface. Furthermore, the representation of the visual field in adjacent low order visual areas is, in a certain sense, “inverted” (the “visual field sign” alternates between adjacent visual areas). Methods: The alternation of the visual field sign may be taken advantage of to delineate, by means of fMRI, the low order visual areas in the individual subject. The fMRI paradigms are then based on the measurement of the functional responses to visual stimuli which cover the visual field in a periodic manner. Typically, sectors rotating around a fixation point or annuli centred on the fixation point and presenting a varying diameter are used as stimuli. One can then establish easily the one-to-one relationship between the coordinates of the receptive fields of the neuronal populations in the visual field and the phases of the BOLD responses. The gradients in the phase maps permit determining the visual field sign and eventually the delineation of the low order visual areas. This talk is aimed as an introduction to the principles of those delineation techniques. References: M.I.Sereno et al, Science 268, 889 (1995); J.Warnking et al, NeuroImage 17, 1665 (2002)

Teaching Session I 8:30 am - 10:00 am

Ruys

Cardiac MRI/MRS: a new frontier 7 Blood flow in and around the heart G. Reiter1, U. Reiter1, H. Mächler2, M. Perthel3, M. Zink4, P. Lyszcz5, R. Rienmüller1; 1General Radiology, University Hospital Graz, Graz, AUSTRIA, 2Surgery, University Hospital Graz, Graz, AUSTRIA, 3Surgery, Herz-Kreislaufzentrum Bad Bevensen, Bad Bevensen, GERMANY, 4Anesthesiology, University Hospital Graz, Graz, AUSTRIA, 5Radioelectronics, Technical University Warsaw, Warsaw, POLAND. Introduction: Assessment of blood flow patterns is of central importance for understanding function and dysfunction of the heart and its surrounding vessels. Being noninvasive on the one hand and providing spatial and velocity information simultaneously on the other hand, magnetic resonance imaging represents a unique technique for this fluid mechanical task. Methods: Whereas cine gradient echo images show blood flow qualitatively, which is helpful for the location of stenoses and regurgation, phase contrast method is able to quantify flow velocities. Because each phase contrast scan, aside from the velocity encoded images also provides anatomical images, flow can be determined via velocity times area of interest. Conventional applications of phase contrast method measure flow velocities in one

S4

Cardiac MRI/MRS: a new frontier

dimension, which is suitable for situations with a preferred blood flow direction. In the heart, however, no such direction exists and fluid dynamics should be assessed three-dimensional. To acquire three-dimensional velocity fields in the left ventricle or even in the total heart, a phase contrast sequence grid protocol can be applied. By post-processing, acquired data can be visualized both as colorencoded vectors on the anatomical phase contrast images and as color-encoded cones in a three-dimensional reconstruction of the left ventricle respectively the heart. Results: One-dimensional phase contrast measurements can be applied to measure cardiac output, to quantify shunts and pulmonary regurgation after tetralogy of Fallot repair, to estimate the severity of stenoses and to determine the flow in the coronary arteries. Studying the three-dimensional flow in the left ventricle of sheep as well as of healthy human volunteers, it turns out, that there are characteristic patterns: During systole blood flow is uniformly directed towards the aorta. In the early diastole blood streaming from the left atrium through the mitral valve is essentially directed apically. Nearby the leaflets of the mitral valve, vortices – fluid structures with circular motion – are formed. Both vortices increase in the later diastole, where the vortex at the anterior leaflet is more pronounced than the one at the posterior leaflet. If the natural mitral valve is replaced by a mechanical mitral valve prosthesis, deviations from the characteristic, normal blood flow patters can be observed, which depend on the prosthesis’ orientation. Conclusion: Magnetic resonance phase contrast method is a useful tool to measure flow-related parameters in and around the heart. The possibility to determine the three-dimensional dynamics of blood enables new cardiovascular questions and answers.

8 Cardiac and coronary imaging: the basics V. Sinitsyn; Tomography Dept., Cardiology Center, Moscow, RUSSIAN FEDERATION. Cardiac and coronary imaging is one of the most interesting fields of MRI. Due to substantial technical improvements, MRI is used more and more widely for imaging of the heart and vessels. Today cardiac MRI can be performed practically with any type of MR imager, but high-field systems with strong gradients have advantages, especially for coronary and functional imaging. A basic requirement in cardiac MRI is reliable synchronization with ECG. Imaging protocol depends on purpose of the study. Manufactures of MRI offer some predefined protocols but they should be adapted according to the type of cardiovascular pathology studied. Choice of pulse sequences and imaging planes strongly depends on the clinical question. Desired appearance of the blood pool (bright/black blood imaging) dictates the choice of pulse sequence. A right balance between spatial and temporal resolution must be observed. SE and TSE sequences are preferred for studies of cardiac morphology, breath-hold fast GE and EPI sequences are used for functional cine-MR and perfusion studies. Segmented k-space data acquisition is preferable for coronary angiography and heart chamber's functional assessment. MRI is a modern “standard of reference” for the quantification of heart chamber's volumes and myocardial mass. MR has excellent possibilities for blood flow quantification. The technique of myocardial tagging provides unique functional information about properties of ventricular wall. There are several techniques for MR coronary angiography, like free-breathing or breath-hold 2D and 3D modalities. Parallel image acquisition techniques (GRAPPA, SENSE, SMASH), steady-state free precession imaging (true-FISP,

FIESTA) allows for faster acquisition of images with better signalto-noise and contrast-to-noise ratios. Real-time cardiac imaging has become feasible. MR contrast agents are used for studies of myocardial perfusion and viability and for enhancement of the blood flow signal in coronary MRI. MRI with contrast enhancement is regarded as a competitor to PET for studies of myocardial viability. Cardiac MRI must be used for the right indications. Evaluation of anatomy or function if questions remain after echocardiography remains the main indication for cardiac MRI. MRI is regarded as a method of choice in the evaluation of congenital heart diseases, pericardial pathology, intra- and pericardial masses, imaging of the right ventricle and pulmonary artery. Several scientific societies (like ESC and SCMR) have offered classifications of indications for cardiac MRI. Technical progress and accumulation of scientific and clinical data will increase demand for cardiac MRI and lead to recognition of new indications for clinical use of this modality.

9 Myocardial spectroscopy – clinical application M. Beer; Institut für Röntgendiagnostik, Universitätsklinik Würzburg, Würzburg, GERMANY. Purpose: MR-spectroscopy is the only method for non-invasive assessment of cardiac metabolism. Especially energy metabolism is vital for regular cardiac function. Recent technical advances like reproducable quantification methods, dedicated data acquisition and postprocessing methods have been introduced. Method: For spatial localisation, 2D and 3D-CSI techniques (chemical shift imaging) are most often used as standard. Comparison with external or internal standards is used for quantification. A novel technique, Spatial Localization with Optimum Pointspread Function (SLOOP), even allows free adaption of voxel borders to cardiac anatomy. Nuclear Overhauser enhancement (NOE) or acquisition-weighting (AW-CSI) can be applied to increase the sensitivity of MRS measurements. Ultra-high field magnets with 2T or 3T have been introduced, increasing the sensitivity of MRS. Results: PCr and ATP are the two main molecules of cardiac energy metabolism. In healthy myocardium, the ratio between PCr and ATP is around 2. In heart failure, this ratio is increasingly reduced. Due to the low spatial resolution of clinical MRS, mostly global heart diseases are studied by MRS. However, improved data acquisition techniques like AW-CSI allow a localized analysis in ischemic heat disease reducing measurement times or increasing spatial resolution (voxel size 16 ccm for 4 different regios of human heart at 1.5T). Determination of absolute values has revealed the combined depletion of PCr and ATP, which can only be detected by quantitative techniques. This improved sensitivity even allows to detect subclinical cardiac commitment in a varity of diseases like multiple sclerosis or different forms of neurodegenerative diseases. The comparison of 1.5T with 2.0T scanner demonstrated an up to 50% increase of signal to noise ratios (SNR) in healthy volunteers. Conclusions: Recent technical advances and the advent of high field magnets may allow MRS to become a clinical routine tool, which offers valuable additional information on cardiac metabolism for detection and assessement of various cardiac diseases.

Abdominal MRI/MRS: new developments

Teaching Session I 8:30 am - 10:00 am

Plate

Abdominal MRI/MRS: new developments 10 Liver imaging B. J. Op de Beeck; Radiology, University Hospital Antwerp UZA, Antwerp, BELGIUM. Magnetic resonance imaging (MRI) of the liver has evolved into a powerful imaging modality for the detection and characterization of focal liver pathology. The advent of fast imaging techniques allows rapid breath-hold whole-liver imaging for lesion detection and characterization. An overview of scanning techniques for a comprehensive MR evaluation will be presented. Use of contrast agents in MRI of the liver has become an indispensable component of a comprehensive scanning protocol. Several classes of MR contrast agents are currently available for clinical use. These include non-specific materials that have an extracellular distribution, materials that are taken up specifically by hepatocytes and excreted in part through the biliary system, and materials that are targeted specifically to the Kupffer cells of the reticuloendothelial system (RES). The properties and indications of each category of contrast agent for liver MRI will be discussed. Benign tumors can arise from each of the cellular components of the liver: hepatocytes, biliary epithelium and mesenchymal tissue. Most benign tumors are discovered incidentally during abdominal ultrasonography, and the most common entities are simple cysts, cavernous hemangiomas, FNH and adenomas. Malignant liver tumors can be divided in primary malignant neoplasms of the liver and secondary malignant liver tumors. The most common primary malignant neoplasm of the liver of hepatocellular origin is hepatocellular carcinoma. The second most common primary malignant neoplasm is of cholangiocellular origin and consists of the intrahepatic cholangiocarcinoma. Secondary malignant liver tumors are metastasis. MRI changes of the liver post-surgery or post-ablation are an increasing challenge for the abdominal radiologists given the increased frequency with which these techniques are performed. Demonstration that malignant disease has limited hepatic involvement may have a substantial impact on patient management. The MR appearance of the most common hepatic lesions will be discussed. Imaging of diffuse liver disease will also be briefly discussed including fatty liver, inflammatory disease, cirrhosis, iron overload and vascular pathologies.

11 Small bowel imaging A. Laghi; Dept. of Radiological Sciences, University of Rome, Rome, ITALY. MR of the small bowel is a field of application which has not been explored for several years and it was confined for a long time within advanced and selected research centers with no clinical impact. Following first two publications dated back in 1985, it is in 1998 and following years that a decisive improvement of scientific contributions was observed. The reason is in the progressive development of adequate technologies, represented by fast imaging. Today, if the solution of major technical problems was obtained,

S5

the optimization of study protocol is still under evaluation. The challenge is between two different methodological approaches: "MR follow-through" following administration of an oral contrast medium; "MR enteroclysis", following administration of contrast medium through a naso-jejunal tube. At the moment several contrast media are available for small bowel study, and they are classified, according to signal intensity, into positive, negative or biphasic. Biphasic agent are those most commonly used for both MR follow-through and MR enteroclysis. Today, the major clinical indication of a study of the small bowel with MR is represented by the evaluation of chronic inflammatory disease (Crohn’s disease). A consistent experience has been obtained in the past years, with the consequent introduction of MR in routine clinical practice in many centres worldwide. Other potential clinical indications, at the moment to be considered as work-inprogress, are the evaluation of malabsorption syndromes (and among them, celiac disease) and assessment of neoplastic diseases.

12 Prostate imaging and spectroscopy A. Heerschap, T. Scheenen, J. Futterer, M. Engelbrecht, F. van Dorsten, J. Barentsz; Radiology, University Medical Center Nijmegen, Nijmegen, NETHERLANDS. Among men above 50 years of age prostate diseases such as cancer or prostatic hyperplasia (BPH) are major health problems Proper diagnosis of prostate cancer is of crucial importance for adequate treatment strategies. However, the presence, location, heterogenous extent and stage of cancer tissue in the prostate are often difficult to determine. Neither PSA levels, digital rectal examination, transrectal ultrasound and conventional T2 weighted MRI are satisfying in this respect. Benign diseases of the prostate, such as prostatitus and benign prostatic hyperplasia (BPH) frequently confound the diagnosis of prostate cancer. Recently, new MR methods have been introduced that can improve the detection of prostate cancer. Most prominent among these are dynamic contrast enhanced MRI (DCE MRI) and 1H MR spectroscopic imaging (1H MRSI). With DCE MRI using Gd-DTPA as contrast agent, the vascularity of prostate tissue can be assessed from its contrast uptake in time; tumor tissue with abnormal blood vessels due to neo-angiogenesis often show

S6

Musculoskeletal MRI/MRS: new developments

more rapid and pronounced enhancement and more rapid washout after the application of Gd DTPA. 1H MRSI allows the spatially resolved detection of metabolites such as citrate and choline, of which the levels change markedly in prostate cancer. In this presentation the potentials and limitations of both these approaches to detect prostate cancer will be discussed. Some key elements in current data acquisition and analysis methods will be described. An overview will be given of results obtained so far and it will be demonstrated that a combination of DCE MRI and 1H MRSI in one examination provides improved diagnostic possibilities. Finally future perspectives of prostate MR at higher field strengths will be indicated. References: 1. Kurhanewicz J, et al. [1996] Radiology 198:795-805 and [2002] JMRI 16, 451; 2. Heerschap A, et al. [1997] Anticancer Res. 17:1455-60; 3. Jager GJ, et al. [1997] Radiology 203:645-52; 4. Hittmair K, et al. [1994] Magn. Reson. Med. 31:567-71; 5. Tofts PS, et al. [1999] J. Magn. Reson. Imag. 10:223-32; 6. Liney GP, et al. [1999] NMR Biomed. 12:39-44. 7. Padhani et al [2001] Radiology 218; 365; 8. Engelbrecht M, et al [2003] Radiology, in press.

Teaching Session II 10:30 am - 12:00 noon

Willem Burger

Musculoskeletal MRI/MRS: new developments 13 Cartilage imaging C. Glaser; Clinical Radiology, Ludwig Maximilians University, Munich, GERMANY. Motivation for the MRI-based assessment of the articular cartilage is the increasing availability of and experience with cartilage dedicated therapeutic options, mainly on the background of acute trauma and/or osteoarthritis. This creates a need for a non-invasive diagnostic tool contributing to establish the need for specific therapy in any individual patient and assisting in the evaluation of the therapeutic efficacy of these new treatment options. Thus, in addition to qualitative, ´visual´ assessment of cartilage, quantitative parameters and data on their validity and reproducibility are required in view of objective follow-up studies. Yet, MRI of the articular cartilage still remains challenging due to its complex composition and small spatial extension. High spacial resolution (HR), good SNR / CNR and reasonable imaging times are prerequisites for an accurate MRI assessment of the articular cartilage. Fat suppression eliminates chemical shift artefacts and improves CNR of the cartilage. Using selective water excitation (WE) contributes to further reduce imaging time. In routine clinical MRI at 1,5T a spatial resolution of 3 x 0.3 x 0.6 mm³ currently is available. T2-w (PD-FS) sequences are considered to be sensitive to focal intracartilaginous signal alterations and show a y good delineation of the cartilage surface from effusion indicating even relatively small cartilage lesions. However, assessment of small lesions is still problematic and suffers from low accuracy. Cartilage volume and thickness constitute simple global and quantitative parameters associated with the development of OA. Several groups have shown the validity of HR 3D MR (mostly T1-w FS GE, 1.5 x 0.3 x 0.3 mm³) sequences with

adequate 3D postprocessing for quantitative cartilage volume and thickness measurements. The technical reproducibility error is small (1 to 6 %, COV) compared to the interindividual variability (15 – 25%) and the range of loss between health and advanced OA (60%). Physical exercise leads to a temporary reduction of patellar cartilage volume of about 6% in young and 3% in elder healthy volunteers and cartilage thickness seems to exhibit diurnal variations in the same order of magnitude. First longitudinal studies indicate an annual rate of loss of 4 – 6% in patients suffering from OA. An emerging focus of interest is the biochemical and structural analysis of the cartilage matrix in intact and defective articular cartilage. Among currently investigated MRI parameters are T2 relaxation and gadolinium uptake mapping, magnetization transfer contrast (MTC) imaging and diffusion weighted imaging.

14 Whole body MRI for metastases S. Eustace, Dr; Radiology, National Orthopaedic & Mater Misericordiae Hospitals, Dublin, IRELAND. This lecture will review developments facilitating clinical whole body MR imaging with an emphasis on the use of moving table top and TurboSTIR tissue excitation. The lecture will review its application in screening for metastases, unknown primary tumour, lymphoma and in the assessment of total tumour burden. The lecture will review practical aspects of image acquisition and subsequent image interpretation, including the use of digital image melding software.

Neuro MRI/MRS: new developments 15 Bone structure H. F. Boehm1,2, T. M. Link1; 1Department of Radiology, TUMuenchen, Munich, GERMANY, 2Institut fuer extraterrestrische Physik, Max-Planck-Society, Garching, GERMANY. Osteoporosis is a metabolic bone disorder characterized by progressive loss of bone mass and deterioration of micro-architecture. The standard diagnostic parameter is bone mineral density (BMD) – a bulk measure for bone mass assessed by dual-energy x-ray absorptiometry (DXA) or quantitative computed tomography (QCT). Osteoporosis affects cortical and trabecular bone resulting in a decreased bio-mechanical strength. Past studies have shown that micro-structure and BMD are independent factors contributing to bone strength. Since in trabecular bone, structural changes usually occur at higher rates than in cortical bone it can be used as an ideal model for the understanding of osteoporosis. Recently, imaging techniques have evolved allowing to depict the trabecular architecture of bone. In this context high-resolution magnetic resonance imaging (HR-MRI) has received considerable attention as a research and clinical tool. HR-MRI is a non-ionizing imaging modality for generating arbitrarily oriented 3D image data. As the MR-signal corresponds to the proton-content of a tissue - due to surrounding marrow and fat - bone is represented as a negative image. Depending on field-strength and the choice of specific pulse-sequences and parameters HR-MRI can be used for invitro as well as in-vivo applications. In human subjects appendicular sites (phalanges, distal radius, calcaneus) have successively been studied using HR-MRI. The image-quality depends on a number of technical factors: trabecular dimensions tend to be overestimated in gradient-echo-sequences as compared to spin-echo-sequences and increase with echo time (TE). Until lately, in-vivo application of spin-echo-sequences was limited due to long examination times and relatively low signal-to-noise ratios. However, novel spin-echo-sequences allowing for very good image quality at reasonable acquisition times have recently been developed. From the image data, micro-structural features obtained by quantitative measures are analysed with respect to the presence of osteoporotic fractures of the spine (in-vivo) or correlated with bio-mechanical strength (in-vitro). Fairly well established are linear structural measures in 2D based on standard histomorphometry (trabecular separation, trabecular volume, connectivity). Further techniques include computation of star-volume, Fourier spectrum, or fractal dimension. Recently, non-linear techniques in 3D have been introduced to the structural analysis of trabecular bone, which are superior to standard measures and BMD in predicting strength and fracture risk. The new parameters are based on the scaling-index-method (SIM), the standard Hough transform, and the Minkowski functionals. Although the results of structural analysis are very promising further development in hard- and software is required to make the new techniques available for clinical routine.

S7

Teaching Session II 10:30 am - 12:00 noon

Jurriaanse

Neuro MRI/MRS: new developments 16 Principles and Applications of Parallel MRI K. P. Prüssmann; Institute for Biomedical Engineering, ETH and University Zurich, Zurich, SWITZERLAND. In the past few years, parallel MRI with receiver coil arrays (1-4) has rapidly advanced from the early research stage to the clinical arena. Enhancing the efficiency of spatial encoding in MRI by twoto approximately eightfold, the parallel approach permits significant scan time reduction in numerous clinically relevant applications. As secondary benefits it equally enables resolution enhancement, artifact suppression, SAR reduction and mitigating acoustic noise. Enhanced encoding efficiency in parallel MRI is the result of exploiting an additional spatial encoding mechanism. Besides conventional gradient encoding, the method relies on the fact that receiver coils of distinct geometry exhibit distinct spatial sensitivity. In fact, coil sensitivity, which is governed by electrodynamics, acts as a genuine encoding mechanism, as it introduces a relative weighting of signal components according to their spatial origin. While mathematically largely equivalent, the two encoding effects differ fundamentally in the underlying physics: Manipulating the spin phase of the object under examination, gradient encoding is necessarily a sequential procedure. Sensitivity encoding, by contrast, exploits coil properties and can thus be performed in parallel by using an array of multiple, simultaneously operated receiver coils. The added information thus acquired permits reducing the number of sequential gradient encoding steps, resulting in faster, more efficient data acquisition. In the technical part, this lecture will cover the basics of parallel imaging technology, including physics, hardware requirements, and specific aspects of image reconstruction and quality. The second part will be a survey of current parallel imaging applications in clinical neuroimaging. References: 1. M. Hutchinson, U. Raff, Magn. Reson. Med. 6, 87-91 (1988) 2. J. B. Ra, C. Y. Rim, Proc. SMRM, 10th Annual Meeting (1991), p. 1240 3. D. K. Sodickson, W. J. Manning, Magn. Reson. Med. 38, 591-603 (1997) 4. K. P. Pruessmann et al., Magn. Reson. Med. 42, 952-962 (1999)

17 From DWI to DTI D. Le Bihan; Shfj/unaf, CEA, Orsay, FRANCE. Molecular diffusion refers to the random translational motion of molecules, which results from the thermal energy carried by these molecules. The success of diffusion MRI is deeply rooted in the powerful concept that, during their diffusion-driven displacements, molecules probe tissue structure at a microscopic scale well beyond the usual image resolution. Water is the most convenient molecular species to study with diffusion MRI. During typical diffusion times of about 50ms, water molecules move in the brain on average over distances around 10 micrometers, bouncing, crossing or interacting

S8

Neuro MRI/MRS: new developments

with many tissue components, such as cell membranes, fibres or macromolecules. The observation of this displacement distribution, on a statistical basis, provides unique clues on the structural features and the geometric organization of neural tissues, as well as changes in those features with physiological or pathological states. Potential clinical applications of water diffusion MRI were suggested very early, but the most successful application since the early 1990s has been brain ischaemia. With its unmatched sensitivity diffusion MRI provides some patients with the opportunity to receive suitable treatment at a stage when brain tissue might still be salvageable. On the other hand, diffusion is truly a three-dimensional process. Hence, molecular mobility in tissues may not be the same in all directions. Diffusion anisotropy in white matter originates from its specific organization in bundles of more or less myelinated axonal fibres running in parallel, although the exact mechanism is still not completely understood: Diffusion in the direction of the fibres is faster than in the perpendicular direction. It appeared quickly that this feature could be exploited to map out the orientation in space of the white matter tracks in the brain. With the introduction of the more rigorous formalism of the Diffusion Tensor MRI, diffusion anisotropy effects could be fully extracted, characterized and exploited, providing even more exquisite details on tissue microstructure. The most advanced application is certainly that of fibre tracking in the brain which, in combination with functional MRI, opens a new window on the important issue of brain connectivity. Diffusion tensor MRI has also appears as a promising tool to look at brain maturation and development, especially to monitor the myelination process, as well as changes in connectivity in relation to disorders, such as dyslexia. Furthermore, recent data suggest that diffusion MRI could also be used to image brain activation by directly visualizing dynamic tissue changes associated with neuronal activation.

(MRS) for improved diagnosis and therapy of patients with brain tumours. Spectra uniquely delineate biochemistry of human tissue in situ. Although MRS gives significantly improved brain tumour categorisation, it is not widely used, partly because radiologists have difficulty in interpreting spectral data. We therefore aimed to develop a user-friendly computer program for spectral classification. Systems development has been informed by (a) a large "training set" of data contributed by members of the consortium and (b) new spectra acquired under agreed protocols. Raw data have been stored in a web-acessible database. Data accrual has taken place under strict quality control procedures at the level of NMR instrumental performance, spectra quality and criteria for acceptance of clinical data. Pattern recognition techniques have been developed for tumour classification together with an intuitive graphical user interface (GUI). The scientific prototypes of the single voxel (SV) and multi voxel (MV) GUI contain, amongst others, three separate "windows": 1. The classification window presents each spectrum as a point in a multidimensional space so that the user can compare an individual spectrum of unknown class with others for which the classification is already known. 2. The image window has two main purposes a)To show the regions in the brain (voxels) from which the spectra are acquired. b) For multivoxel prototypes, to present a “nosologic image” of the brain showing the classification of the different tissues within and surrounding the tumour. 3. The spectrum window displays individual spectra in standard format (x-y plots) together with mean and standard deviation, as well as spectra of different tumour types (figure 1). An industrial prototype containing more than 300 cases from the INTERPRET database is presently at the CE marking stage. It allows automated processing and display of short (TE 20-33 msec) and long (TE 135-136 msec) MRS data with classifiers based in Linear Discriminant Analysis (LDA) of selected heights in short TE spectra. The system has been optimized for the discrimination among aggressive (glioblastoma+metastases), low grade glial and meningioma, although data from 12 major tumour types, abscesses, normal brain and “rare” tumours can be displayed and compared with new data. Scientific protoypes for MV at short and long echo time have also been developed and may merge in the future into a single DSS for human brain tumour diagnosis. Figure: Snapshot of scientific SV-GUI

18 Taking a virtual biopsy of cerebral tumours with 1H MRS: The INTERPRET project .. INTERPRET1, C. Arús2; 1http://carbon.uab.es/INTERPRET/ interpret_partners.shtml, Vth Framework Programme, Cerdanyola del Vallès, SPAIN, 2Departament de Bioquímica i Biologia Molecular, Universitat Autònoma de Barcelona, Cerdanyola del Valles, SPAIN. The objective of INTERPRET (http://carbon.uab.es/INTERPRET/) has been to facilitate the use of magnetic resonance spectroscopy

Parallel Imaging: Basic and Clinical Applications

Teaching Session II 10:30 am - 12:00 noon

Ruys

Parallel Imaging: Basic and Clinical Applications 19 Surface coils, arrays and the benefits of array imaging M. A. Griswold; Lehrstuhl für Experimentelle Physik 5 Biophysik, Universität Würzburg, Würzburg, GERMANY. Unlike many other imaging modalities, MRI is fundamentally limited by the signal to noise ratio (SNR) achievable in a given imaging time. This fact was one of the primary reasons that MRI was so slow in its early stages of development, since many averages were needed to obtain a single image. The first step towards optimizing the SNR came with the introduction of surface coils. These coils only receive signal and noise from a localized region, so that the geometry, and therefore the SNR, can be optimized for a given target object. The only drawback of these surface coils is that optimized surface coils only have a limited range of sensitivity, which greatly reduces their potential in imaging applications such as spinal imaging where a large FOV needs to be covered. The quantum leap in this area was the introduction of multicoil, or phased, arrays. These arrays contain many independent surface coils, each of which can be optimized for SNR for a particular part of the FOV, which in most cases provides substantial gains in SNR compared to volume coils or individual surface coils. These large gains in SNR were one of the fundamental reasons for the success of modern fast gradient systems, since the faster gradients would have been useless without more SNR. More recently the field of parallel imaging has been developed which uses the intrinsic spatial information from array coils to partially encode the image. This represents another substantial acceleration of MR imaging, which comes at the cost of reduced SNR. However, with the further design of optimized imaging arrays, it is anticipated that high accelerations can be achieved with enough SNR to be clinical useful.

20 Parallel MRI for beginners D. J. Lomas; Radiology, University of Cambridge, Cambridge, UNITED KINGDOM. Since the development of in vivo MRI there have been numerous attempts to reduce image acquisition times. Initially this was achieved by more efficient transmit and receive systems, and improved coil designs. More recently the development of clinically useful single and multi-shot sequences on high performance gradient sub-systems have contributed to the achievement of sub-second imaging. This has been particularly useful for imaging the heart, abdomen and pelvis. These approaches are starting to reach the practical limits imposed by the safety constraints regarding the speed of gradient switching and the RF power deposition. Parallel imaging techniques offer an alternative approach to speeding up data acquisition by using spatial sensitivity information to substitute for the phase encoding process used in traditional spin warp image acquisition.

S9

This presentation will introduce in simple descriptive terms the following topics – Basics of Spin Warp Image Acquisition & Aliasing Segmentation and Multi-shot or Echo-train Acquisition Coil Spatial Sensitivity What is needed for Parallel Imaging What are SMASH and SENSE What are the limitations What are the benefits

21 Clinical implications of parallel MRI C. Kuhl; Department of Radiology, University of Bonn, Bonn, GERMANY. Virtually all existing MR imaging techniques require both, a high temporal and a high spatial resolution for optimum detection and classification of disease states. Up to now, increasing gradient strength has been the only strategy to meet the increasing demands of advanced diagnostic imaging applications. Yet, this strategy is limited by physical, economical, and medical considerations: Increasing gradient strength is technically difficult, it is associated with significant hardware costs, and goes along with the risk to induce unwanted side effects such as peripheral neuro-stimulation. With the advent of parallel imaging techniques like SENSE, this dilemma can be overcome. The SENSE-mediated reduction of acquisition time can be invested in three ways: To improve spatial resolution at a given imaging time, or as well cut down aquisition time at a given spatial resolution; further, it can be used to reduce echo time length in order to avoid susceptibility artifacts. For obvious reasons, SENSE is most useful to enhance pulse sequences with high inherent SNR that are acquired with only one signal average. Accordingly, ideal applications are contrast enhanced 3D MR angiography both, for the cerebral as well as abdominal or peripheral vascular system. Furthermore, dynamic contrast-enhanced studies for liver or breast tumors benefit considerably from the increase in spatial resolution that is achievable without any expense on temporal resolution. For many patients, in particular the actually sick ones or for childern, the shorter breath hold times that are feasible with SENSE-enhanced pulse sequences in virtually all abdominal or cardiac applications are vitally important – often enough, patients are too sick to comply with “regular” breathhold regimen. If SENSE is avialable, these patients can, for the first time, be examined with MRI. In sumary, SENSE improves existing pulse sequences – this translates into a more accurate MR diagnosis of vascular diseases or tumors. In addition, it helps in particualr the sick or the small patients to actually benefit from the advantages of state-of-the-art MR imaging techniques. Owing to its stability and reliability, and, of course, owing to its inherent advantages for many MR imaging applications, SENSE is now fully integrated into routine clinical practice. This talk is meant to provide the audience with an overview on current SENSE applications in our Department.

S10

Image Interpretation for Physicists

Teaching Session II 10:30 am - 12:00 noon

Plate

Image Interpretation for Physicists 22 Image interpretation for physicists: brain and spine Ö. Özsarlak; Department of Radiology, University Hospital Antwerp, Antwerp, BELGIUM. Magnetic resonance imaging (MRI) has revolutionized radiological investigation and has been especially important in neuroradiology. The basic of the image interpretation is that the pathology alters in size, shape and location. Additionally, structural and functional changes can also be observed on MRI. The major role of the medical imaging, particularly MRI is picturing pathology. There are two essential elements to be considered: geometry and contrast of the lesion. The geometry, namely physical property varies by region (anatomy) and by function (physiology). Image contrast depends on imaging technique and alters by contrast agent administration. In this respect, MRI has the advantage of studying lesions on their T1 and T2 values. Water, fat, bony structures, gray matter, white matter, cerebrospinal fluid, acute, subacute or old blood can be differentiated easily depending on their signal intensities in various imaging sequences. Combination of the study protocols, using fat suppression techniques and contrast (gadolinium) studies will help to diagnose or differentiate certain lesions, or lesion categories from others. First of all appropriate knowledge of brain and spine anatomy is required at the starting point of image interpretation. Intraaxial, extraaxial, and bony anatomy of the brain and spine should be well understood in infants, children and adults. The associated signs and characteristics of lesions, such as involvement pattern, distribution, edema, relation to the cerebrospinal fluid, mass effect, and finally contrast enhancement patterns will help for narrowing the differential diagnosis. MRI has also advantages to study vascular structures. For the spinal lesions, three major locations should be distinguished; vertebral body, disc space, and spinal canal. The alignement, shape and the signal intensity are the major criteria in the interpretation of vertebral body, whereas the height, signal intensity and the contour for the disc space.

23 Image interpretation for physicists: abdomen D. Vanbeckevoort, F. Claus, D. Bielen, S. Dymarkowski, E. Nijs, K. Op de Beeck, G. Marchal; Radiology, University Hospitals Leuven, Leuven, BELGIUM. Despite its initial promise, MRI has been less commonly applied in the abdomen than in other anatomic areas such as brain and spine. In fact, the abdominal cavity contains multiple organs and structures of different size, function and complexity. Moreover, different types of motion (e.g. respiratory excursions and bowel peristalsis) are a challenging topic for imaging. During the past few years however, MRI technology available for imaging of the abdomen has rapidly evolved and a large range of new applications has been introduced. Abdominal MRI is no longer reserved for further characterisation of diseases initially imaged by ultrasound or CT, but currently also used as a first choice examination. A wide range of sequences characterises MRI and a profound knowledge and meticulous application of the appropriate technique

are necessary to obtain morphological and/or functional information. The use of high-performance gradient systems and localised phased-array body coils enables to cover large volumes of interest and to obtain a high spatial resolution within a reasonable acquisition time, bypassing many of the motion artefacts that previously posed limitations. Because of the introduction of a wide variety of new contrast agents, MRI has become an important tool in detecting and characterising focal lesions in the liver, pancreas and spleen, as well in small structures such as the adrenal glands or lymph nodes. The optimisation of sequences such as single-shot rapid acquisition with relaxation enhancement (RARE) has led to the implementation of high-quality MR cholangiopancreaticography (MRCP) which is becoming an important non-invasive modality for assessment of the hepatobiliary system. The potential of MRI to provide dynamic information is intriguing, e.g. dynamic MR defaecography is used to image functional disorders of the pelvic floor in proctology. Shortened acquisition times in magnetic resonance angiography and the use of (non)breathhold imaging have reduced the impact of motion artefacts, and have made it possible to obtain successive images in the arterial and portal phase. In short, MRI plays an important role in the detection and characterisation of a wide range of abdominal diseases. Its use is no longer restricted to cases where other investigations are inconclusive. New advances in hardware and software have increased the clinical utility of MRI and have led to reasonable examination times.

24 Image interpretation for physicists: musculo-skeletal J. L. M. A. Gielen; Department of Radiology, University and University Hospital of Antwerp, Antwerp, BELGIUM. The purpose of this presentation is to illustrate gross musculoskeletal anatomy using a virtual model of a joint. The various components of the joint will be discussed, including: bone, cartilage, ligaments, tendons and muscles, menisci, and articular fluid. Keeping in mind that function is reflected in logical anatomical relationships, general rules can be worked out that are easily applied. For each component, semiology of musculoskeletal pathology and the appropriate MR imaging techniques to evaluate them will be discussed, such as the choice of coil (surface coil, volume coil, phased-array coils), imaging planes, sequence choice, slice thickness, parameter settings, use of parallel imaging techniques, value of Gd-enhanced sequences, use of magnetization transfer contrast, ... New developments in sequence design and 'hot topics' will be addressed such as kinematical MRI in orthopaedics and sports medicine, 3-D evaluation of the structure and density of bone, the effect of high magnetic field strength and bandwidth on imaging of biomedical implants, whole body MRI for tumor staging; differentiation between benign and malignant compression fractures of the spine measuring apparent diffusion coefficient of vertebral body bone marrow and strategies to minimize RF inhomogeneity and chemical shifts artefacts….. In conclusion, we hope that this presentation may contribute to a better understanding of muskuloskeletal anatomy and image interpretation for physicists. An in-depth knowledge of the clinical issues in muskuloskeletal MR imaging will lead to an improved collaboration between basic scientists and clinical radiologists, and will help to focus further developments in problem-oriented research.

Opening Lecture / MR Imaging of Angiogenesis and Microvasculature Thursday, September 18, 2003

Opening Lecture 12:45 pm - 1:25 pm

Willem Burger

Opening Lecture 25 Two decades of MRI: impact on basic science and clinical radiology in a University environment A. Baert; Department of Radiology, U.Z. Sint Barbara, Leuven, BELGIUM. Basic research has benefited from continuous developments in methodology and techniques, carried out in both university and manufacturers laboratories (such as gradient hardware, sequence development, parallel imaging) Majot advances were booked in fields which benefited from the non-invasive character of MR such as fMRI in the human which has become a major tool in neuroscience, spectroscopy and imaging in the study of human diseases using transgenic mice, in the development of molecular imaging and drug targeting. The main expectations from the clinical introduction of MRI two decades ago were: better tissue characterization and higher specificity based on a higher intrinsic contrast resolution of the image, replacement of RX based examinations by a method based on nonionising radiation and decrease in the use of iodinated contrast media and increased use of new MR contrast media with the same high specificity as the tracers used in NM but providing higher spatial resolution. The main concerns about the possible limitations of MRI as a clinical method were related to: some patient “unfriendly“ aspects such as the difficulties in using life support and monitoring equipment during the MR study, difficulties in applying the methods in infants and children and claustrophobia in adult patients, the costs and the complexity of the siting requirements, the limited spatial resolution, poor visualization of moving organs and possible, yet unknown, nocious effects such as specific absorption and heating effects. Immense and rapid technical progress has overcome many if not all of these concerns and limitations. The clinical impact of MR imaging has been tremendous and many expectations have indeed been fulfilled or even exceeded particularly in the central nervous system and spine, the musculoskeletal system, in imaging of the abdominal organs, the heart and the peripheral vascular system. The reasons why clinical MRI has not replaced to a larger extent conventional X-ray and CT are frequently not related to pure scientific and medical arguments but are largely due to regulation and budget restraints. On the other hand the clinical impact of new specific MR contrast media has been disappointing to some extent. Two decades following the introduction of MRI as a clinical and research modality technical progress is still quite rapid and the future of this method looks bright. New sequences and newly designed RF coils are being developed constantly. The potential of increasing field strength is being explored. New applications in perfusion studies of the myocard, kidney, brain and in malignant tumors without the use of contrast media (spin labeling) can be expected. Progress can be expected in the study of myocard motility and

S11

stress as well as in exploration of the lung ventilation using hyperpolarized gases. Furthermore MR may play a major role in the exact placement and detection of stem cells in target organs. Molecular imaging and spectroscopy may benefit from higher field strengths. MRI may also gain in importance for the monitoring of results of molecular therapy by new MR tracers.

Plenary Session 1:30 pm - 3:00 pm

Willem Burger

MR Imaging of Angiogenesis and Microvasculature 26 MRI of cerebral microvasculature R. Turner; Wellcome Department of Imaging Neuroscience, Institute of Neurology, London, UNITED KINGDOM. The draining vein problem is recognized as one of the most severe constraints on the spatial resolution of BOLD contrast fMRI, used widely in imaging neuroscience. Changes in blood oxygenation arising from local brain activity-related changes in blood flow propagate downstream in veins and can give rise to spurious activation at sites remote from neuronal activity. The geometry of the venous vasculature is quite regular in structure and is well depicted in photomicrographs. Quantitative analysis of this geometry, together with hydrodynamic considerations, permit upper bounds dependent on the area of cortical neuronal activity to be derived for the spatial extent of draining vein contamination. It has been estimated that an activated cortical area of 100 mm2 will generate an oxygenation change in venous blood that extends without dilution along the vein no more than about 4 mm beyond the edge of the activated area. At greater distances along the draining vein this oxygenation change will be diluted. The model leads to a quantitative prediction of the functional form of this dilution. On the basis of this calculation upper limits to the spatial resolution of BOLD contrast fMRI have been estimated. These vary according to the area of cortical neurons activated, giving rise to the recommendation that fMRI experiments should be designed in such a way that the differential activation of interest between conditions will be limited in extent, thus minimizing the draining vein effect. In order to provide experimental support for these estimates it is desirable to obtain more detailed MR images of cerebral microvasculature. At high magnetic fields, such as 4.7 T, it is possible to obtain voxel dimensions of 0.5 x 0.4 x 0.4 mm3 dimensions, which is sufficient to depict much of the pial vasculature. Use of strong T2* weighting, achieved using the relatively long echo time of 15 ms (following the work of Reichenbach) provides good quality venograms that can be used as the basis for detailed modelling of BOLD effects downstream.

27 MRI of tumoral neoangiogenesis – preclinical M. Neeman; Biological Regulation, The Weizmann Institute of Science, Rehovot, ISRAEL. Growth of blood vessels is an integral component in progression of solid tumors. Blood vessels not only provide the growing tumor with essential nutrients and help remove waste products, but also provide routes for tumor cell escape and transport to distant loca-

S12

Functional MRI: Methodology

tions during metastatic dissemination. Non-invasive imaging tools in general, and MRI in particular, are important for probing the structure, function and temporal and spatial remodeling of tumor blood vessels. Imaging methods can assist in basic research aimed to unravel the regulation of angiogenesis, in development of drugs for suppressing vascular expansion and in development of novel approaches for following patients during therapy. There are multiple ways by which contrast in MRI can be sensitized to angiogenesis, and each contrast mechanism provides a glimpse of a different stage in the angiogenic cascade. Intrinsic contrast methods can be used to probe blood flow and oxygenation. Administration of contrast agents in preclinical studies can mimic clinical procedures using the same contrast agents, or alternatively can be used for development of novel, frequently macromolecular, contrast material. These methods can be used for mapping vessel diameter, blood flow, volume fraction, and vessel permeability. Specifically designed targeted contrast agents can be used for imaging particular molecules that are over-expressed in neovasculature. Recent studies applied MRI for monitoring angiogenesis in a range of preclinical models in combination with administration of inducers and inhibitors of angiogenesis, along with models designed for altered expression of specific growth factors. For example MRI contributed to the study of modulation of expression and activity of VEGF and its impact on vessel permeability, recruitment of perivascular support, vascular co-option and the role of angiogenesis and VEGF in metastatic spread. Changes in some of these parameters during tumor growth and during treatment could be monitored in the same animals. In particular changes in vascular permeability proved to be very sensitive to pharmacologic inhibition of VEGF. In summary, MRI is increasingly used for evaluation of alteration in angiogenesis in pre-clinical models with altered expression of key angiogenic pathways, and in evaluation of novel therapeutic approaches that aim to suppress tumor angiogenesis or to destroy the existing vascular bed. Methods developed either for basic research or for drug evaluation, form the basis for clinical imaging of tumor angiogenesis for aiding in diagnosis, prognosis, response to therapy and patient monitoring.

28 MRI of tumoral neoangiogenesis – clinical A. Jackson, Professor; Imaging Science & BIomedical Engineering, University of Manchester, Manchester, UNITED KINGDOM. The development of new blood vessels is an essential requirement for growth of many pathological tissues, particularly tumours. The mechanisms which promote angiogenesis appear to be generic so that therapeutic approaches which target the process may be expected to have broad application. This has resulted in extensive research to develop methods to quantify the angiogenic process and the related features of the microvasculature. Because of the spatial variation in microvascular structure and the need for repeated measurements imaging based techniques are favored.Dynamic contrast enhanced MRI (DCE-MRI) and more recently CT provide temporal data which allows the calculation of physiological variables related to the microvasculature. These include regional blood volume (rBV), Blood flow (BF) and endothelial permeability surface area product (PS). Each of these parameters has potential theoretical benefits in describing specific features of the microvasculature. This talk will describe the range of available pa-

rameters and their likely biological correlates. We will review and compare the approaches to image analysis used to calculate them and will review the results of studies to apply these techniques in a range of clinical scenarios.

Scientific Session 3:30 pm - 5:10 pm

Willem Burger

Functional MRI: Methodology 29 Mapping the effects of magnetic field inhomogeneity in EPI for fMRI P. Mannfolk1, D. van Westen2, E. M. Larsson2, J. Olsrud2; 1Department of Radiation Physics, Lund University, Lund, SWEDEN, 2Department of Diagnostic Radiology, Lund University, Lund, SWEDEN. Purpose/Introduction: Geometrical distortion and signal drop in EPI due to local magnetic field inhomogeneity varies within the brain, between individuals and between MR systems. In fMRI this affects coregistration with anatomical images and can also affect statistical power in group analyses (1). The purpose of this study was to map distortion in EPI of the brain, which is relevant to planning and analysis of fMRI. Subjects and Methods: fMRI was performed in 5 healthy volunteers using a 3T head scanner (Siemens Magnetom Allegra). A gradient echo sequence (TE = 4.9 and 7.4 ms) was used to obtain field maps and functional data were acquired using single shot EPI (64x64 matrix, 3mm isotropic voxels, TE/TR = 30/3000ms). A computer program was written in IDL (Research Systems Inc) to calculate the spatial distortion (mm) for each voxel. Original phase difference maps were unwarped and scaled to represent field inhomogeneity and further converted to displacements in the phase encoding direction (2). For each subject, an EPI volume was normalized to the EPI template in SPM99. The resulting transformation was applied to the displacement maps and mean displacement- and standard deviation maps were calculated (n=5). Results: Figures show normalised mean displacement maps (a) and standard deviation-maps (b). In the skull base, displacements as large as 10 mm are found as well as large intersubject variations (1SD < 5mm). More cranially, for example in Broca's area and the sensorimotor cortex, displacements are generally smaller than 3mm with 1SD < 2mm. Discussion/Conclusion: Distortion in EPI due to local magnetic field inhomogeneity was quantified for a group of healthy volunteers. The relatively small mean displacement (less than one voxel) and intersubject variations in most parts of the brain introduce little error when coregistering EPI and anatomical images, but could possibly affect fMRI group analyses (1). In regions with large displacements or intersubject variation, distortion correction should certainly be considered. Mapping the effect of magnetic field inhomogeneity into normalized space (SPM99-template) makes it possible to relate the need of correction to the localisation of functional areas. References: 1. Cusack R, Brett M, Osswald K [2003] Neuroimage 18:127-142 2. Hutton C, Bork A, Josephs O, Deichmann R, Ashburner J, Turner R [2002] Neuroimage 16:217-240

Functional MRI: Methodology

S13

References: 1. Weisskoff,Proc. SMRM(1992) 2. Jezzard,MRM 34(1995) 3. Reber,MRM 39(1998) 4. Cusack,NeuroImage(2003)

30 A multiple echo approach for robust correction of geometrical distortions in echo planar imaging C. Windischberger1, A. Rauscher1, M. Barth2, E. Moser1,2; 1Institute of Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Department of Radiodiagnostics, University of Vienna, Vienna, AUSTRIA. Introduction: Echo-planar images suffer from geometric distortions due to susceptibility differences at tissue boundaries. The amount of displacement is proportional to the shift in the local magnetic field, and using field maps it is possible to correct for these effects. The estimation of field maps, however, is not trivial, as the phase derived from the NMR data is limited to values between -π and +π. Phase wrapping occurs and makes a straight-forward field map estimation by phase difference calculation impossible. Here we present a simple algebraic approach that allows fast and reliable assessment of the magnetic field distribution. Subjects and Methods: The principle is to derive the field distribution not from a single phase difference image alone, but using several phase images acquired at different echo times. Here, we use phase maps from three images. For each pixel we construct seven sets of phase values, assuming phase wraps either between the first two or last two echoes or both. The linear regression over TE and the chi^2-error function is calculated for all sets. The slope with the minimal fitting error is then proportional to the B0 in the respective pixel. Displacement maps are calculated from the B0-values and the reconstructed images are shifted accordingly. Measurements were performed on a 3 Tesla Medspec S300 (Bruker Biospin, Germany) using a FLASH-sequence with 125kHz readout bandwidth, 30 slices, TR=450ms and TEs of 6, 6.5 and 7.5ms, respectively. EPI data sets were acquired with MA=128x128 and TE=46ms. Results: The figure shows the results for a single slice: (a) original FLASH image, (b) raw phase image at TE=6ms, (c) calculated B0map, (d) distorted B0-map to correct EPI data, (e) original EPI, (f) distortion-corrected EPI slice. White lines are overlaid to the EPI slices for better visualization of the distortion-correction result. Discussion: The method presented allows fast and robust estimation of the magnetic field distribution in the human brain. It requires only three minutes of additional data acquisition and uses a simple algorithm that can be implemented very easily, still covering field differences of up to 1000 Hz with the parameters used here. Hence, it may help to make geometric correction a standard post-processing procedure, improving anatomical localization in functional MRI. Acknowledgements: This work has been supported by the Austrian National Bank (ÖNB-JF9305) and the Hochschuljubiläumsstiftung of the City of Vienna (1472/2002).

31 Increased heterogeneity of fMRI due to antihistamine S. Bells, M. D. Noseworthy; Diagnostic Imaging, The Hospital for Sick Children, Toronto, ON, CANADA. Purpose/Introduction: A great deal of heterogeneity exists in fMRI data. Even within the same subject, results on successive days or scan sessions often differ in the number of significantly activated pixels and/or intensity of activation. Recently we have shown that the ingestion of a fatty meal decreases BOLD contrast demonstrating one source of signal variability within fMRI measurements [1]. Since antihistamines are known vasomodulators we investigated whether their ingestion could contribute to within and between subject fMRI data heterogeneity. Subjects and Methods: Using fMRI motor cortex activation was investigated in 3 healthy fasted right handed subjects (two male) before and 1hr post-ingestion of chlorpheniramine (4mg). The motor paradigm consisted of bilateral finger tapping (2Hz) with 6 blocks of 30s/30s off/on phases. Data was acquired using a GE LX 1.5 T MRI using a gradient echo sequence with a spiral readout (24 slices, 5mm thick, TE/TR/α=40/2000/90, 1 NEX, 64x64 matrix).

S14

Functional MRI: Methodology

Analysis was performed using AFNI [2]. Results: Chlorpheniramine ingestion resulted in a decrease in BOLD contrast and a reduction in the number of significantly activated voxels in the primary motor cortex and supplementary motor cortex (P<1.0x10-12) (Fig.1) The temporal data from each subject correlated well with the hemodynamic wave function before and after ingestion of antihistamine, with the greatest correlation and intensity before ingestion.

Fig.1. Bilateral finger tapping significantly decreased BOLD activation following administration of chlorpheniramine (4 mg) (pre ingestion=left : post ingestion=right). Discussion / Conclusion: Antihistamines interact with histamine receptors and thereby affect physiological functions such as cognition, movement, and arousal [3][4][5]. Histamine is an important in situ vasomodulator, changes in which can lead to altered microvascular perfusion and blood volume. From these results we suggest ingestion of antihistamine such as chlorpheniramine can contribute to between and within subject heterogeneity. Therefore subjects should be questioned about their use of such common pharmaceuticals prior to fMRI evaluation. References: 1. Noseworthy et al. (2003). Hum. Brain Mapping (accepted) 2. Cox (1996) Comp. Biomed. Res. 29:162-173 3. Kreitel et al. (2002) Neuroscienc.114: 935-943 4. Mochizuki et al. (2002). Hum. Psychopharm.. 17:431-418 5. Okamura et al. (2000). Brit. J. Pharm. 129:115-123

32 A reproducibility study on functional connectivity as identified by low frequency temporal blood oxygenation level-dependent fluctuations J. G. Hirsch1, M. J. Lowe2, S. Schwenk1, C. Rossmanith1, M. G. Hennerici1, A. Gass1; 1NMR Research Neurology, Dept of Neurology, University Hospital Mannheim, Mannheim, GERMANY, 2Division of Radiology, The Cleveland Clinic, Cleveland, OH. Introduction: In functionally related regions of the brain, synchronised fluctuations of cerebral blood flow have been observed. Experimental studies suggest that these fluctuations are likely to reflect oscillations of neuronal activity [1]. As BOLD MRI is thought to represent neuronal activity by neurovascular coupling it has been accepted, that reproducibility of fMRI studies is limited due to numerous variables involved in the underlying physiology. We investigated the reproducibility of resting state fcMRI studies comparing fcMRI maps obtained in two controls at five time points. Materials/Methods: Timeseries of BOLD-weighted data were acquired on a 3.0T Siemens Allegra using an EPI-FID sequence (TE=30 ms, TH=6mm, MAT=64x64, FOV=240x240 mm2). For the symmetric blockdesign fMR study (alternating 32s rest/stimulation), 15 slices were collected (TR=2000ms); for the resting and

continuously activated state fcMR studies, 3 slices were acquired (TR=250ms). Two different paradigms were chosen: A) simultaneously bilateral finger-tapping, B) auditory stimulation (backwardspeech).Repeated scans were done 5 times within 48 hours. Careful positioning assured reproducible slice positions monitored on T1w. T2*w images were Hamming-filtered [2]; fcMR data were digitally low-pass-filtered (<0.08 Hz) to remove temporal fluctuations arising from cardiac and respiratory-related physiological noise [3]. Blockdesign fMRI identified activated areas in right and left precentral gyrus (PMC), SMA, basal ganglia and cerebellar regions with paradigm A, and insular gyri and the medial geniculate with paradigm B.

For cross-correlation estimation in fcMR timeseries, reference ROIs were drawn in right PMC and in right Heschl-gyrus. Connectivity maps were calculated showing the corrected crosscorrelation between each single pixel and the reference ROI [4].

Results: Both motor and auditory stimulation showed robust and reproducible results in regard to the anatomic localisation and number of activated pixels. Correlation analysis of LFBF showed in a total of 20 maps matching spatial distributions of synchronous LFBF in the presumed motor network located in both PMCs extending up to midline cortex. Discussion/Conclusion: In an attempt to evaluate reproducibility of LFBF correlations it was demonstrated that by adjusting the statisitical criteria a similar anatomical pattern was identified in the motor and auditory network regardless of the continuous activation or resting state. As LFBF may be influenced by the underlying basal neuronal activity, one would assume that different brain states might influence the reproducibility of fcMRI. 1. Leopold D, Soc Neuroscience [2002] 32nd Ann Meeting, #325.7. 2. Lowe MJ [1997] Magn Res Med 37:723. 3. Lowe MJ [1998] Neurolmage 7:119. 4. Lowe MJ [2000] Proc ISMRM 8:799.

33 FMRI signal changes in temporal region related to EEG rhythmic delta oscillations during hyperventilation induced hypocapnia M. J. Mäkiranta1, J. Ruohonen, M.Sc.1, K. Suominen, Ph.Lic.2, E. Sonkajärvi, MD3, T. Salomäki, MD, Ph.D.3, V. Kiviniemi, MD1, S. Alahuhta, MD, Ph.D.3, V. Jäntti, MD, Ph.D.4, O. Tervonen, MD, PhD1; 1Department of Diagnostic Radiology, University of Oulu, Oulu, FINLAND, 2Department of Clinical Neurophysiology, Oulu University Hospital, Oulu, FINLAND, 3Department of Anesthesiology, University of Oulu, Oulu, FINLAND, 4Ragnar Granit Institute, Tampere University of Technology, Tampere, FINLAND.

Functional MRI: Methodology Purpose: This study explores the question: does there exist any differences in hyperventilation (HV) induced BOLD signal changes between two groups 1) with and 2) without intermittent rhythmic delta activation (IRDA) in EEG? In EEG, over 2 minutes of hyperventilation induces intermittent rhythmic oscillations of 22.5Hz in children and some young subjects. In previous fMRI studies, the existence of this phenomenon has not been acknowledged. IRDA is of special interest also because frontal or temporal rhythmic delta can be a sign of epilepsy or even localize the epileptic brain tissue (Di Gennaro et al., 2003). Subjects and Methods: Four subjects (21-24 yrs, non-smoking, right handed, 3 female, 1 male) screened with EEG to have IRDA during the last minute of 3min HV and their age-matched controls, who did not have this reaction, were studied. HV (3 min) was performed during 8min fMRI (GRE-EPI, TE=60ms or TE=40ms, TR 6000ms, i.e., 6 axial slices/1 second, acquisition delay 5000ms due to simultaneous EEG recording) in 1.5T scanner. Physiological parameters were collected with anesthesia monitor using pulse oximeter and end-tidal CO2 measurement. The magneto-compatible EEG-equipment was used for continuous EEG recording (fs=1000Hz). IRDA power (integration of FFT, 4-second window, i.e., 4000 samples) versus time was calculated from derivations Fp1-Cz, Fp2-Cz, O1-Fz and O2-Fz, during acquisition delay. Anatomical ROIs were segmented from two axial slices and average BOLD response of the segment was calculated. Results: Individual data about degree of hypocapnia showed tendency to be slightly lighter and the recovery faster in the control group than in the IRDA group. In fMRI ROI analysis, only group average signal change of temporal gray matter differed prominently between the groups. During IRDA power change, this difference vanished (Fig. 1). Conclusions: It is concluded that an identifiable change in BOLD signal intensity in temporal area was associated with intermittent rhythmic delta activation due to hyperventilation induced hypocapnia in healthy volunteers. This model can serve as a clearly defined EEG pattern related to increased excitation for BOLD contrast fMRI studies, with further relevance in serious brain pathologies, like tumors and epilepsy. Reference Di Gennaro et al. [2003] Clin Neurophysiol 114:70-8. FIG. 1.

S15

34 Electrical forepaw stimulation with physiologically relevant frequencies yields optimal Functional MRI data N. Van Camp1, M. Verhoye1, A. Volny-Luraghi2, E. De Schutter2, A. Van der Linden1; 1Bio-Imaging Lab, University of Antwerp, Antwerpen, BELGIUM, 2Laboratory for Theoretical Neurobiology, University of Antwerp, Antwerpen, BELGIUM. Introduction: Despite the fact that 7-8Hz frequencies are physiologically relevant in the rat’s somatosensory system (8Hz is the main whisking frequency and 7-8Hz oscillations may reflect the baseline dynamic state of the rat’s somatosensory system), combined fMRI and electrophysiological studies during electrical forepaw stimulation of the rat demonstrated decreasing BOLD-signal changes and SEP-amplitudes at stimulation frequencies above 3Hzi. It was our aim to investigate the effect of different stimulation frequencies and pulse-widths on the activation in the somatosensory cortex. Methods: The fMRI study was performed on male adult Wistar rats (n=12,350g), anaesthetized with a single bolus injection of alfa-chloralose (60mg/kg,IP). Images were obtained on a SMIS MR microscope (MRRS) with a 7T horizontal bore magnet and 8cm aperture self-shielded gradients with strength of 0.1T/m (Oxford Instruments), using a surface receiving- and helmholtz transmitting-coil. We used a heavily T2*-weighted single-slice EPI sequence (Slice thickness:2mm, 1mm anterior to the bregma, FOV:35mm, acquisition-matrix:64x64, TR/TE:1sec/26msec). For stimulation of the forepaw, silver needle electrodes were placed between the toes of the rat and connected to the stimulator (WPI stimulator, Sarasota Florida USA) with isolated cupper wires. Square electrical pulses were applied in two different stimulation-paradigms: First different frequencies (1-5-7-8-10-12 Hz) were used with an amplitude of 1mA and pulse width of 10msec (8rats); secondly different pulse widths (0.3-10-50 msec) were tested at a frequency of 8Hz (4rats). Data were processed by Medx-software (Version 3.41, Sensor Systems, Inc, Sterling, USA). Results: All stimulation frequencies resulted in significant BOLD activations in the forelimb region of the contralateral somatosensory cortex. The activated area varied significantly with the applied stimulation frequency and was significantly larger at frequencies around 8Hz. Signal intensity changes at frequencies around 8 Hz were also observed to be much higher. The same observations were obtained for a stimulus pulse width of 10msec, as compared to 0.3msec or 50msec. Discussion: In contrast to what others have demonstrated1, we surprisingly observed the largest BOLD-signal changes at frequencies around 8Hz. Unlike these studies, which used a pulsewidth of 0.3msec1, we used a pulse-width of 10msec. This demonstrated that not only stimulation frequency and intensity of a stimulus2, but also the stimulus pulse-width is important. Moreover it appears that similar to the observation that whisker stimulation at whisking frequency3 provided optimal fMRI data, application of physiologically relevantstimulation frequencies for somatosensory stimulation also yielded the best fMRI results. Endnotes: i Brinker G. et al.(1999)MRM41: 469-473 ii Spenger et al,(2000)Exp Neurology166,246-253 iii Yang et al.(1996) PNAS93, 475-478

S16

Functional MRI: Methodology

35 Brain activity associated with successful and unsuccessful response stopping: Inhibition and error processing A. Chevrier1, R. Schachar2, M. D. Noseworthy1; 1Diagnostic Imaging, The Hospital for Sick Children, Toronto, ON, CANADA, 2Psychiatry, The Hospital for Sick Children, Toronto, ON, CANADA. Purpose/Introduction: Inhibitory motor control is essential to executive function and affects development, normal behavior, and psychopathology. Inhibition, in disorders such as attention deficit hyperactivity disorder (ADHD), has been studied using the well documented Stop Signal Task (SST). No imaging study has yet examined whole-brain event-related activity during SST performance. As such we performed an event-related fMRI study of healthy adults during the SST to assess inhibition and error processing. Subjects and Methods: Six subjects performed 300 trials of the SST, a two choice reaction time task where subjects respond to a visual stimulus ("X" right, "O" left hand button response). On 33% of trials, background color changed to red (250 ms post-stimulus), signifying a stop trial. The stop delay adapted to responses such that # successful (SI) = # unsuccessful (SR) trials. Stop delay increased/decreased by 50 ms after SR/SI trials (inter-trial interval, ITI = [2.5;3.5] seconds, 6s rest between blocks). Imaging was performed with a GE LX 1.5T scanner, using a gradient echo sequence with a spiral readout (TE/TR/α = 40/2000/90, 24 slices, 6mm thick, FOV = 20 cm, 4096 points along 1 spiral-out interleaf) 1. All events were recorded with time of occurrence. A general linear model of stimulus functions convolved with the hemodynamic response function (HRF) was used. Baseline, linear drift, and 6 point HRF’s for each condition were estimated. Intensity maps were calculated from area under HRF, warped into Talairach, and blurred (5mm FWHM). ANOVA was performed with event types as levels, converted to t-maps according to their degrees of freedom, and corrected for multiple comparisons in accordance with Gaussian field theory such that the overall α<0.05.

Results: Bilateral activation during successful inhibition was observed in the anterior cingulate and caudate, and in right motor areas (fig.1). During failed inhibition medial-frontal and posterior cingulate showed deactivations proximal to middle frontal activation (fig. 2). Discussion/Conclusion: The contrast (SI – response execution) revealed that the caudate and cingulate are the best candidates for the source, and motor area 6 the site of inhibition (Figure 1). The contrast (SR- response execution) revealed no positive activations. The SR activation map showed medial-frontal deactivation proximal to middle-frontal activation, implying that the ERN (or Ne) observed on SR trials could be due to the withdrawal of a destructively interfering neural field generator (fig.2). 1. Glover and Lee (1995) MRM 46:515:522.

36 Application of the neural gas algorithm in fMRI analysis E. Dimitriadou1, M. Barth2, K. Hornik1, E. Moser3,2; 1Statistics, University of Technology, Vienna, AUSTRIA, 2Dept. of Radiodiagnostics, University and General Hospital, Vienna, AUSTRIA, 3Dept. of Medical Physics, University, Vienna, AUSTRIA. Introduction: Clustering methods are often used as exploratory data analysis (EDA) tools in fMRI. Those methods have their pros and cons as they deal different with spatial patterns, data distributions and noise characteristics. High performance and stability is important for an algorithm. We present a clustering method, namely neural gas (ngas), as an on-line, winner-take-most approach [1]. It is compared to its corresponding winner-takes-all method (hardcl) [1] and winner-takes-all batch method (kmeans) [2]. The latter is broadly used in fMRI clustering [3, 4, 5, 6] where the update of only the winning center (winner-takes-all) is performed before any adaptations are done (batch-mode). Subjects and Methods: Different simulation designs were used to cover a broad spectrum of data set characteristics (6 different noise characteristics - gaussian (t) and in-vivo (h) noise with 3 different CNRs of 1.33, 1.66, and 2 each (t4-t6 and h4-h6), 3 different cluster initializations of 5, 10, and 20 INC's, 50 simulation runs each). Neural gas is an example of a soft competitive learning algorithm, where for every input pattern the corresponding reference vectors (centers) are updated based on a rank order according to their distance to the input pattern. Then all centers are adapted according to their rank. Quantitative performance assessment was done comparing TPs and FPs for all methods. Results: The figure shows box-and-whisker plots for FP and TP dispersion for all data sets and methods, using INC=20. Discussion: From the simulation results, it is obvious that neural gas performs better in comparison to its competitor, concerning both number of TPs and FPs and their dispersion over the runs. However, we found that ngas performs not optimal with small INCs (5 or 10). This was expected, as center updating is not conservative like in the case of the other methods. In general, ngas decreases the dependency on random initializations, and the system is able to get out of poor local minima more easily. Overall, neural gas proves to be the most stable and robust method when reasonably high INCs are used. Acknowledgements: This study is funded by Austrian National Bank (OeNB-Project-No.9201).

Sequences and Techniques: Parallel Imaging References: 1. T. Martinetz et al, IEEE Tran. on NN 4, p.558-569(1993). 2. Y. Linde et al, IEEE Tran. on Communications 1, p.84-95(1980). 3. C. Goutte et al, HBM 13, p.165-183(2001). 4. H. Fischer, Magn.Reson.Med. 41, p.124-131(1999). 5. U. Moeller et al, NeuroImage 17, p.431-446(2002). 6. A. Baune et al, NeuroImage 9, p.477-489(1999)

S17

moved using GRAPPA. Conclusions: It has been shown that it is possible to derive weights which allow reconstruction of data that are shifted relative to the acquired data. This method is particularly advantageous for dynamic imaging, since the difficult inverse process only needs to be done once for each series, while conjugate gradient [1] methods must perform the entire process again for each additional frame. The application of this method to other symmetric trajectories (spiral) should also be possible and should result in more dramatic improvements. References: 1. Pruessmann,MRM46:638-651(2001) 2. Griswold,MRM47(6):1202-1210(2002)

Scientific Session 3:30 pm - 5:10 pm

Jurriaanse

Sequences and Techniques: Parallel Imaging 37 Direct parallel imaging reconstruction of radially sampled data using GRAPPA M. A. Griswold, R. M. Heidemann, P. M. Jakob; Lehrstuhl für Experimentelle Physik 5 - Biophysik, Universität Würzburg, Würzburg, GERMANY. Introduction: In many cases, non-Cartesian sampling trajectories offer distinct advantages for MRI compared to normal Cartesian sampling. However, the application of parallel imaging to nonCartesian trajectories is nontrivial, requiring the solution of large systems of equations [1]. Here we demonstrate that in many cases, such as projection reconstruction (PR), this process can be simplified using the GRAPPA reconstruction [2]. Theory: The GRAPPA method reconstructs lines in each coil by forming combinations of neighboring lines. In the case of normal GRAPPA, the weights for this combination are optimized for shifting data in k-space by a specific amount in a specific direction. However, one can also derive weights which shift data by a specific amount, but in a direction relative to the other lines by providing the appropriate reference data. If one examines PR sampling, it can be seen that all points at a particular radius have the same relative displacement with respect to neighboring lines. A relative shift operator can be derived by rearranging all of the projections from a reference dataset followed by a normal GRAPPA reconstruction in strips along readout. The similarity of this approach to normal Cartesian sampling should be obvious. All missing projections can then be reconstructed in a time comparable to several GRAPPA reconstructions. Results: Fig2 shows two invivo scans obtained with an 8 channel array. The top, left image shows the 32(proj)x256(r.o.) normal reconstruction, while the right shows the 2x GRAPPA reconstruction (64x256). The bottom row shows an example from swallowing, however in this case the GRAPPA reconstruction is 3x resulting in a 96x256 image. As can be seen, the aliasing artifacts are largely re-

38 The GRAPPA Operator M. A. Griswold, R. M. Heidemann, P. M. Jakob; Lehrstuhl für Experimentelle Physik 5 - Biophysik, Universität Würzburg, Würzburg, GERMANY. Introduction: GRAPPA (1) uses multiple lines of k-space data to reconstruct missing k-space lines in each coil of the array. In this abstract, we demonstrate that several additional relationships can be derived. It is shown that the GRAPPA reconstruction can be

S18

Sequences and Techniques: Parallel Imaging

reformulated as a matrix operator. Using this formalism, we show that there exists an infinitesimal GRAPPA operator which can shift data by arbitrarily small amounts. Other desired k-space shifts can then be accomplished through repeated applications of this operator. Theory: Given signal Sj(k)at position k in k-space in each coil of the array j=1...N for N coils, we can derive a set of weights G1which shift the data in each coil to a position k+1: Sj(k+1)=G1Sj(k) If we then desire a shift to k+2, we would normally derive another set of weights G2which correspond to a shift of +2: Sj(k+2)=G2Sj(k) This could also be rewritten as: Sj(k+2)=G2Sj(k)=G1Sj(k+1)=G1G1Sj(k) This implies that the weights satisfy the following condition: G2=G1G1 This can be generalized to an infinitesimal shift of d: Sj(k+d)=GSj(k) Other shifts can be derived by multiple applications of the weights, G, which we refer to as the GRAPPA operator. Also, since a shift of +1 followed by a shift of -1 should result in the same data as the input: Sj(k)=G1G-1Sj(k)then G1G-1=I This implies that the weights for a +1 shift are the inverse of the weights for a -1 shift. With these relationships in hand, all desired shifts can be derived from a single fit operation followed by repeated applications of the same weights. Results: Figure1 shows an example from a 4coil array. The direct fits for d=0.1,0.01,0.001 are all essentially identical. Additionally reconstructed data can be "recycled" and used for reconstructing other lines. Figure2(left) shows a 4x accelerated image which was then reconstructed by 2x(middle). These data were then used for another 2x reconstruction, resulting in a total acceleration of 4. In this case, half of the lines that were reconstructed in the first step were used as source lines in the second. Conclusions: The GRAPPA fit process results in an operator which shifts data in k-space. This results in several interesting relationships between the weights. Besides using these relationships for normal reconstructions, new schemes can be derived. Error correction can be implemented. Finally, the reconstruction of nonCartesian data is easier since the inverse problem is substantially reduced. References: 1. Griswold,MRM47(6):1202-1210(2002)

39 Sensitivity encoding at 7 Tesla using transceive microstrip coil arrays F. Wiesinger1, P. Van de Moortele2, G. Adriany2, P. Boesiger1, K. Ugurbil2, K. P. Pruessmann1; 1Institute for Biomedical Engineering, ETH and University Zurich, Zurich, SWITZERLAND, 2Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN. Introduction: Ultra-high field strength and parallel imaging are topics of ongoing technological development in MRI. The inherent advantages and disadvantages of the two concepts are largely contrarian: parallel acquisition enables scan time reduction, artefact suppression and even SAR reduction, yet at the expense of SNR. High field provides increased basic SNR but enhances field inhomogeneity and SAR issues. Furthermore, theory predicts that shorter wavelength at high field should enhance parallel imaging performance (1). Thus the two approaches are in fact highly complementary, making their combination very promising. In the present work we report the implementation and application of the parallel SENSE (SENSitivity Encoding) technique (2) at 7 Tesla. Methods: The SENSE technique was integrated with a 7 Tesla whole-body magnet (Magnex), equipped with a Varian Inova console and Siemens Symphony gradient amplifiers. Data were acquired with custom-built 4- and 8-channel transceive stripline head coil arrays. A prerequisite for reliable SENSE reconstruction is accurate coil sensitivity calibration. With a homogeneous reference coil unavailable at 7 T, a sum-of-squares combination was used as a denominator for creating raw-sensitivity maps (S1, Figure 1). The remaining severe phase distortions due to enhanced B0 inhomogeneity, chemical shift and wavelength effects were eliminated by estimating a phase pattern common to all array elements (3). The phase-corrected raw sensitivity maps (S2) were extrapolated and smoothed by a variational approach (S3). Results: Figure 2 shows SENSE brain images obtained in a healthy volunteer with the 4- (top) and the 8-coil array (bottom), using SENSE reduction factors R between 1 and 6. The values of R and the resulting mean geometry factor g, which reflects noise enhancement, are shown above each image. The 4-channel data were obtained with a T1-weighted inversion recovery turbo FLASH sequence (FOV=210x170mm2, matrix=256x256, full acquisition time 102.4s). The 8-channel data were acquired using a FLASH sequence (FOV=240x220mm2, matrix=256x336, full acquisition time 94.1s).

Sequences and Techniques: Parallel Imaging

Discussion: These results demonstrate that SENSE is feasible at 7 Tesla if specific challenges of sensitivity calibration are addressed. In agreement with theoretical predictions (1) the geometry factors found are significantly lower than with similar setups at 1.5 and 3 Tesla, indicating a crucial role of wavelength effects. Concerns that dielectric resonance may prevent efficient sensitivity encoding by rendering coil sensitivities too similar could not be confirmed. References: 1. Wiesinger et al, Proc 10th ISMRM (2002) p.191. 2. Pruessmann et al, MRM 42:952-962 (1999). 3. de Zwart et al, MRM 48: 1011-1020 (2002).

40 Controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA) F. Breuer, M. Blaimer, M. Müller, R. Heidemann, M. Griswold, P. Jakob; Universität Würzburg, Physikalisches Institut , EP5, Würzburg, GERMANY. Introduction: A new approach to improve image quality in partially parallel MRI is presented. Aliasing artifacts are manipulated by modifications during the data acquisition period in order to achieve improved conditions for image reconstruction. In contrast to conventional parallel imaging methods, folded images with essentially identical coil sensitivities can be separated by using adapted, reordered coil sensitivity information, according to the modifications during the acquisition period. Theory and Methods: While standard partially parallel acquisition (PPA) strategies are entirely concentrated on optimizing the post processing strategy, CAIPIRINHA generates controlled aliasing patterns using an additional spatial modulation of the magnetization through modified radio frequency (RF) excitation and gradient encoding schemes. With this, an overlap of areas with highly different coil sensitivities can be achieved, resulting in significantly improved unfolding conditions, especially if the individual slices have almost the same coil sensitivities. The technique works with all standard PPA-algorithms by adapting and reordering coil sensitivity maps or k-space lines, respectively. As an example, the acquisition scheme for multi-slice imaging using a simultaneous excitation of two slices is shown in Figure 1. Individual slices are shifted by a fraction of the field of view (FOV) with respect to each other by applying a different RF-phase cycle to each slice. Another example for controlled aliasing is a modification of the gradient tables in 3D-encoding (see Figure 2). Experiments were performed on a 1.5 T Vision scanner (Siemens, Erlangen, Germany) with a 4-channel receiver body array. All image reconstructions were done in Matlab (The Mathworks, Natick, MA, USA) using conventional PPA algorithms and adapted coil sensitivity information. Results: As shown in Figure 3, CAIPIRINHA in multi-slice imaging provides an artifact free slice separation (geometry factor close

S19

to one at every location) even with directly adjacent slices, while using a conventional multi-slice SENSE [1,2] approach, image slices with almost identical coil sensitivities cannot be separated. Discussion: In conclusion, CAIPIRINHA provides a new degree of freedom in partially parallel MRI by modifying aliasing artifacts during data acquisition in a controlled manner. Thus, CAIPIRINHA has the potential to exploit the variation of coil sensitivities along multiple dimensions in a more efficient manner than standard PPA techniques. Therefore an improved image quality and higher scan time reduction can be achieved. Acknowledgements: This work was funded by the DFG 827/4-1. References: 1. Larkman DJ et al. JMRI 2001 Feb;13(2):313-7 2. Weiger M et al. MAGMA 2002; 14:10-19

S20

Sequences and Techniques: Parallel Imaging ently higher SNR, but also allows reduction of the echotime of a spin-echo measurement without loss in spectral resolution on the ppm scale. The most promising future application of ultrafast MRSI might be 3D whole brain high resolution MRSI within a potentially clinically manageable time. References: 1. Dydak, U. et al., MRM 46:713-722, 2001. 2. Duyn, J.H. et al., MRM 30:409-414, 1993. 3. Dydak, U. et al., MRM, in press.

Figure 1

41 High resolution MRSI in less than a minute U. Dydak1, D. Meier1, R. Lamerichs2, P. Boesiger1; 1Institute for Biomedical Engineering, University and ETH Zürich, Zürich, SWITZERLAND, 2Philips Medical Systems, Philips, Best, NETHERLANDS. Introduction: Using the higher SNR and the higher spectral resolution available at 3T makes MR spectroscopic imaging (MRSI) possible in less than one minute. Parallel MRSI (SENSE-SI) [1] combined with a multi spin-echo MRSI technique (TSI) [2] has proven successful at 1.5T, but limited to an echotime (TE) of 288 ms and an echo train length of two in order to keep the SNR of the spectra reasonable [3]. It is shown here, that the higher spectral resolution at 3T allows reduction of the echotime to 144 ms, thus allowing for echo train lengths as high as six (TSI6) and a total scan time reduction as high as a factor of 12. Methods: MRSI data were acquired from a single slice (20 mm) from a healthy brain (TR=1500ms, TE=144ms, BW=2250Hz, 256 samples , FOV= 230mm, 24x24 voxel). All experiments were performed on a 3T Philips Intera whole body system, using six acquisition channels from a phased-array headcoil (MRI Devices Corp.). For SENSE measurements, data was undersampled in both spatial dimensions by a factor of two, thus obtaining a SENSE reduction factor of four. One to six spin-echoes were acquired per TR in respective measurements, allowing for comparison of measurements with different echo train lengths. Postprocessing included spatial and spectral filtering, B0-correction and additional water suppression. The absolute signal of several metabolite peaks were summed over a 0.2 ppm interval each and metabolite images were interpolated to 2562 pixels. Results: While conventional MRSI (24x24 matrix) lasted 11 minutes, it was possible to acquire the same MRSI data within only 54 s by using SENSE combined with a spin-echo train length of six. Creatine images as well as a sample spectrum from the in vivo measurement are shown for conventional MRSI (Fig.1a) and SENSE-TSI6 (Fig.1b), demonstrating that the loss in SNR is within acceptable limits. Conclusion: While SNR considerations remain a critical factor for fast MRSI, higher magnetic field strength not only yields an inher-

42 Auto-SENSE cine cardiac imaging H. Köstler, M. Beer, C. Ritter, D. Hahn, J. Sandstede; Institut für Röntgendiagnostik, Universität Würzburg, Würzburg, GERMANY. Purpose: Cardiac MR cine imaging during breath hold is a compromise between spatial and temporal resolution and duration of breath hold. Parallel imaging techniques allow the acquisition of images of unchanged resolution compared to Fourier imaging in reduced scan time. Aim of this work was to implement the AutoSENSE technique for cine imaging of the human heart and to test this technique in healthy volunteers. Subjects and Methods: Auto-SENSE cine imaging was implemented on a 1.5 T scanner equipped with a 4 element body phased array coil. In Auto-SENSE alternate reduced k-spaces are acquired in analogy to the UNFOLD technique [1]. From a complete k-space the relative maps of the coil sensitivities can be calculated. These maps, which are independent of the heart phase, allow the SENSE reconstruction of full FOV images from the reduced k-spaces. The imaging parameters of the used FLASH sequence were TR 10 ms, TE 4.8 ms, acquisition window per phase 90 ms, delay between reconstructed images by view sharing 50 ms, flip angle 30°, FOV 286 mm x 380 mm, spatial resolution 2.27 mm x 1.48 mm, slice thickness 8 mm, acquisition time 8 heart beats. Results: Auto-SENSE cine imaging could be performed successfully in all volunteers. No additional temporal filtering was necessary [2]. Neither a pre scan nor additional lines in k-space are required to generate the sensitivity maps. Compared to conventional Fourier imaging the duration of the breath hold could be reduced from 16 to 8 heart beats. Conclusion: The reduced acquisition time of Auto-SENSE cine imaging will be advantageous especially for sick patients who have problems to hold their breath. A combination of Auto-SENSE with 3D – trueFISP imaging [3] may facilitate cine imaging in future. Literature: 1. Madore B, Glover GH, Pelc NJ. Unaliasing by Fourier encoding the overlaps using the temporal dimension (UNFOLD), applied to cardiac imaging and fMRI, Magn. Reson Med. 42, 813-828 (1999)

Sequences and Techniques: Parallel Imaging 2. Kellman P, Epstein FH, McVeigh E. Adaptive Sensitivity Encoding Incorporating Temporal Filtering (TSENSE), Magn Reson Med 45 , 846 - 852 (2001) 3. Jung BA, Hennig J, Scheffler K. Single-breathhold 3DtrueFISP cine cardiac imaging, Magn Reson Med 48 , 921925 (2002)

43 Parallel acquisition technique and multiaverage imaging of the spine Ö. Özsarlak, J. W. Van Goethem, P. M. Parizel, A. M. De Schepper; Department of Radiology, University Hospital Antwerp, Antwerp, BELGIUM. Purpose: To evaluate the effect of multiaverage spine imaging in uncooperative patients. Methods and Materials: Ten patients were examined with sagittal TSE T2-weighted images, with following parameters: TR: 2900ms, TE: 102ms, slice thickness: 3mm, 512 matrix and with PAT factors 1 to 4. PATx1 acquisition performed in 167 sec and with 2 averages, PATx2 acquisition in 178 sec and with 4 averages, PATx3 acquisition in 164 sec and with 5 averages, and PATx4 acquisition in 161 sec and with 6 averages. The same acquisitions were repeated, and patients were asked to swallow twice and cough once during each acquisition at a certain time. SNR were measured and compared from the bone, disc and medullary structures. Results: By using the PATx2 there was a 43% increase in SNR in comparison to that of PATx1, and a 5% increase with PATx3 in vertebral bone. For the disc space, there was a 28% increase in SNR by PATx2. For the cervical medulla, 47% and 15% increase rates were calculated in SNR by applying PAT factors 3 and 2, repectively. Increased signal-to-noise ratios were resulted in decreased motion and swallow artifacts during acquisitions. Conclusion: Multiaverage imaging combined with parallel acquisition technique allowed SNR improvements in patients with motion simulation during the acquisition. This technique, that has no additional increase in acquisition time can be applied to uncooperative patients.

S21

44 Fast volumetric MRI with a SENSE/GRAPPA-hybrid M. Blaimer, F. Breuer, M. A. Griswold, R. M. Heidemann, M. Müller, P. M. Jakob; Physikalisches Institut, Universität Würzburg, Würzburg, GERMANY. Introduction: A new concept for fast volumetric MRI using a combination of Sensitivity Encoding (SENSE, [1]) and Generalized Autocalibrating Partially Parallel Acquisitions (GRAPPA, [2]) is described. As described in [3] an improved image quality can be achieved with reduced phase encoding (PE) in two spatial dimensions. The computational approach presented here allows the use of a standard reconstruction algorithm for 2D imaging and thereby minimizes the computational effort. Theory and Methods: For image reconstruction the undersampled k-space data is Fourier-transformed along the partition encoding (PAE) direction. Therefore a hybrid data space is generated, since the PAE dimension is transformed into the image domain, while the other dimensions remain in k-space. Because of the undersampling along the PAE direction, aliasing occurs and partitions from multiple sections from the excited slab are superimposed (This is equivalent to SENSE). To unfold the aliased partitions, their corresponding low resolution reference images are reordered into a new matrix (see Figure 1). This matrix is Fourier-transformed to calculate the reconstruction parameters as in conventional GRAPPA. Finally a standard GRAPPA reconstruction is applied to obtain the separated full resolution full field-of-view (FOV) images. Experiments were performed on a 1.5 T Siemens Quantum Symphony Scanner (Siemens Medical Systems, Erlangen, Germany) using an 8-channel head coil (MRI Devices, Waukesha, WI). A gradient echo sequence was used to acquire a 3D data set. Image reconstructions were done with Matlab (The Mathworks, Natick, MA, USA). Results: A reconstructed partition of the decimated 3D-data set (total reduction factor R = RPAE x RPE = 6) with good reconstruction quality is shown in Figure 2. Discussion: The SENSE/GRAPPA-hybrid shows improved image quality compared to conventional parallel imaging strategies which use data reduction in only one PE direction. Therefore higher scan time reduction can be achieved. Furthermore the computational approach shown here allows the use of standard reconstruction algorithms for 2D imaging since the 3D problem is transformed into multiple 2D problems. Acknowledgements: This work was funded by the DFG 827/4-1. References: 1. Pruessmann KP, et al., Magn Reson Med 1999; 42:952-962. 2. Griswold MA, et al., Magn Reson Med 2002; 47:1202-1210. 3. Weiger M, et al., MAGMA 2002; 14:10-19.

S22

Human Brain MRS: Paediatrics, Metabolic Disease, Focal Infections and Post Mortem Basal ganglia

Scientific Session 3:30 pm - 5:10 pm

Ruys

Human Brain MRS: Paediatrics, Metabolic Disease, Focal Infections and Post Mortem 45 Prognostic value of 1H MRS and ADC in severe perinatal asphyxia of full-term neonates P. M. Walker1, C. Boichot1, A. Lalande1, C. Durand2, J. Gouyon3, F. Brunotte1; 1Service de Spectroscopie RMN, University Hospital, Dijon, FRANCE, 2Service de Radiologie, University Hospital, Dijon, FRANCE, 3Service de Pédiatrie, University Hospital, Dijon, FRANCE. Introduction: Perinatal asphyxia may result in hypoxic-ischaemic encephalopathy (HIE) leading to cerebral palsy and mental retardation in surviving neonates. In addition to clinical criteria, MRI represents an important modality for the prognostic evaluation of perinatal asphyxia. However, early conventional MRI may be hindered by the presence of diffuse œdema, often observed in neonates irrespective of the ultimate clinical outcome. The present study was undertaken to correlate quantitative MRI/MRS with the clinical outcome of neonates with severe perinatal asphyxia. Methods: Twenty full-term neonates of gestational age 38-40 weeks were clinically diagnosed with severe perinatal asphyxia (Sarnat ≥ 2). The MRI and MRS examinations were performed within 8 days of admission. Neonates with congenital malformations, inherited metabolic disorders and infection were excluded from this retrospective study. All infants were clinically reevaluated for neurodevelopment up to 24 months. Clinical outcome was classified as favourable, indicating normal evolution or only minor handicap and poor, indicating death or major neurological handicap. Spectra were acquired on a 1.5 T Magnetom Vision using CSI sequences (PRESS TE 270 ms / TE 80 ms). Quantitative data were acquired on metabolite concentrations, relaxation times (T1 and T2) and ADC in the basal ganglia, in frontal and parieto-occipital white matter. An unpaired t-test was used to compare the outcome groups. Statistical significance was accepted for p < 0.05. Results: The neonates were matched in age and weight at the time of examination: the delay between initial diagnosis and exploration being similar. Only data from the basal ganglia provided significant differences between the two groups. Although similar trends were observed in frontal and parieto-occipital white matter data, no statistical differences were noted – the quality of spectra being somewhat poorer in these locations. No relaxation time differences were noted.

Normal or minor handicap (n = 9)

Died or severe handicap (n = 11)

Gestational age (days) 275 ± 8 Weight (kg) 2.86 ± 0.61 Delay MRI/MRS (days) 4.4 ± 1.8

277 ± 12 3.15 ± 0.42 5.9 ± 3.4

Cho (mM) Cr (mM) NAA (mM)

2.88 ± 0.40 * 6.14 ± 0.57 * 4.48 ± 0.12 **

2.38 ± 0.24 5.26 ± 1.04 2.66 ± 0.77

Lac/Cho

0.16 ± 0.16 *

0.47 ± 0.36

T1 (ms) T2 (ms) ADC (10-5cm²/s)

1533 ± 104 157 ± 8 1.23 ± 0.09 *

1570 ± 100 169 ± 24 0.86 ± 0.29

* p < 0.01 ** p < 0.001

Discussion: The most pertinent results were significant reductions in [NAA] and ADC in the poor outcome group vs favourable outcome group, the reduction in ADC suggesting the presence of cytotoxic œdema in basal ganglia. The Lac/Cho ratio was also elevated in the poor outcome group, confirming previous studies. However, the non-significant results in relaxation times suggest conventional MRI, as used in our work, lacks specificity in the territories studied, diffuse œdema being present in 15/20 neonates. Our data suggest that MRS derived biochemical markers and tissue ADC from basal ganglia are useful in the evaluation of severe perinatal asphyxia and in the identification of neonates at risk.

46 Possible metabolic signature of kernicterus identified using 1H MRS W. Oakden1, S. Blaser1,2, A. Moore3, M. D. Noseworthy1,4; 1Diagnostic Imaging, The Hospital for Sick Children, Toronto, ON, CANADA, 2Medical Imaging, The University of Toronto, Toronto, ON, CANADA, 3Neurology, The Hospital for Sick Children, Toronto, ON, CANADA, 4Medical Biophysics, The University of Toronto, Toronto, ON, CANADA. Purpose/Introduction: Kernicterus is the clinical term for neurological disorders resulting from the precipitation of unconjugated bilirubin in the brain. In some patients, especially pre-term infants, even relatively low levels of bilirubin can produce kernicterus [1]. Identification is qualitative and typically difficult, based solely on

Human Brain MRS: Paediatrics, Metabolic Disease, Focal Infections and Post Mortem imaging studies where there are abnormal signal changes in various brain regions including the globus pallidus [1,2] and thalamus [2]. Retrospective analysis of proton magnetic resonance spectroscopy (1H-MRS) data from basal ganglia of patients with either hyperbilirubinemia or symptoms of kernicterus was done to see if a metabolic pattern of this disorder could be identified. Subjects and Methods: A total of 7 patients between the ages of 3 days and 3 weeks, with either hyperbilirubinemia or symptoms of kernicterus, were scanned using a GE Signa 1.5T LX MRI scanner (General Electric, Milwaukee, WI). 1H-MRS was done using a single voxel PRESS sequence (TE=144, TR=1500ms, 256 scans, 6-8cm3), centered in the basal ganglia. Quantitative analysis was done using LCModel [3] and between subject comparison was done through calculation of metabolite ratios relative to creatine (Cr). Results: Relative to Cr, levels of taurine (0.79±0.23), glutamate/glutamine (1.26±0.30), and inositol (2.50±1.1) were found to be higher than normal for this age. The glutamate level exceeded that of glutamine (0.73±0.14 vs. 0.52±0.27). NAA levels were not appreciably unusual for this age group (1.17±0.25). Discussion/Conclusion: Several animal studies have suggested that one of the effects of bilirubin in the brain is to modify Nmethyl-D-aspartate (NMDA) receptors [4] which are also involved in the release of taurine under excitotoxic conditions [5]. Taurine is an inhibitory neurotransmitter which is hypothesized to moderate the potentially toxic effects of glutamate [5]. Increased glutamate has been also observed in ischemia studies [6], along with a parallel increase in taurine [5]. Ischemia, however, causes a decrease in NAA [7] which was not observed here. We suggest therefore that brain changes due to kernicterus may be reflected in an increase in both taurine and glutamate. 1. Sugama et al. (2001) Ped. Neurol. 25:328-331. 2. Yilmaz and Eckinci (2002) Eur. Radiol. 12:1837-1839. 3. Provencher (1993) MRM 30:672-679. 4. Grojean et al. (2001) Pediatr Res. 49:507-513. 5. Saransaari and Oja (2000) Amino Acids 19:509-526. 6. Nishizawa (2001) Life Sci. 69:369-381. 7. Rutgers et al. (2000) JMRI 11:279-286.

47 Detection and quantitation of brain phenylalanine in healthy subjects by 1H MRS: A feasibility study R. Kreis1, T. Lutz2, K. Zwygart1, G. F. Hoffmann2, C. Boesch1, D. Rating3, J. Pietz3; 1Department of Clinical Research, University Berne, Berne, SWITZERLAND, 2Department of General Pediatrics, University Heidelberg, Heidelberg, GERMANY, 3Department of Pediatric Neurology, University Heidelberg, Heidelberg, GERMANY. Introduction: In patients with phenylketonuria (PKU), 1H MRS has been used to determine cerebral concentrations of phenylalanine (Phe) and its blood-brain-barrier (BBB) transport ((1-4) and refs therein). In healthy subjects, brain Phe concentrations (~50µM) are too low to be recorded by 1H-MR spectroscopy (MRS) and BBB kinetics cannot be determined. Quantitative aspects of BBB-kinetics and steady-state blood-brain ratios for Phe in PKU are heavily disputed (1,3). Hence, it would be of great interest to establish the methodology to determine brain Phe also in healthy subjects in order to compare PKU results to normal BBB dynamics. In an exploratory study, single oral Phe-loads were applied to investigate the ensuing course of blood Phe. Repetitive Phe loading was administered in one subject to test the feasibility of brain Phe detection by 1H-MRS in healthy subjects.

S23

Methods: Three healthy subjects were given a single oral L-Phe load (100 mg/kgBW) and subsequent Phe plasma levels were observed. One healthy subject was given an initial (125 mg/kgBW) and 4 subsequent (35 mg/kgBW at 1h-intervals) Phe loads. Brain Phe was measured by 1H-MRS before and during the load from a 70 cm3 cerebral volume. Technical settings have been described earlier (4). Data processing included the use of a reference lineshape, parameterization of all signals and model-fitting using priorknowledge constraints (4). Quantitation was based on a 4-compartment brain model and an assumed water content.

Results and Discussion: After a single Phe dose plasma Phe peaked at ~1 mM within 1 h after intake and declined steadily afterwards (N=3). In contrast, repetitive Phe intake led to a plateau of ~1.3 mM plasma Phe (N=1). These sustained high plasma values gave rise to continued Phe inflow into the brain enabling the observation (Fig. 1) and quantitation (Fig. 2) of brain Phe in a healthy subject. The blood/brain ratio at the end of the observation period approached a value of 4 - similar to what was found in PKU patients at steady-state (1,4). From the good definition of brain Phe levels (Fig. 2) it can be expected that this setup will enable reliable investigations of BBB dynamics for Phe also in healthy subjects. References: 1. Kreis R. Eur J Pediatr 2000;159(Suppl 2):S126-S128. 2. Weglage J, et al. Ann Neurol 2001;50:463-467. 3. Pietz J, et al. Ann Neurol 2002;52:382-383. 4. Rupp A, et al. J Cereb Blood Flow Metab 2001;21:276-284.

Acknowledgement: Supported by the Swiss National Foundation (31-059082)

S24

Human Brain MRS: Paediatrics, Metabolic Disease, Focal Infections and Post Mortem

48 Proton MR Spectroscopy in Autosomal Dominant Adult-Onset Leucodystrophy (ADLD) L. Hallberg1, J. Weis1, A. Melberg2, R. Raininko1; 1Department of Radiology, Uppsala university, Uppsala, SWEDEN, 2Department of Neurology, Uppsala university, Uppsala, SWEDEN. Introduction: Autosomal dominant adult-onset leukodystrophy (ADLD) is a slowly progressive disorder affecting both infra and supratentorial white matter. Clinically it presents in the fourth and fifth decades of life with autonomic function loss followed by a decline of fine motor skills. However, there may exist extensive white matter changes in subjects who have not yet presented any clinical symptoms. A survival rate of 20 years is common. MRI findings have been described in one American-Irish (1), one Italian (2) and two Swedish families (3) until now. We now describe MRS findings in the Swedish families. Subjects and Methods: Ten subjects (5 M, 5 F), aged 34-61 years, who had ADLD (6 subjects) or were descendants of such a patient (4 subjects) were studied clinically and with brain MRI and 1H MRS. An individually adjusted voxel (13-23 ml) was placed in the frontoparietal white matter (Fig.1). The study was performed at 1.5T using a PRESS sequence. A pilot study was made in four subjects, aged 24-61 years, using TR/TE 2000/32,136 ms. Six subjects, aged 34-58, TR/TE was 6000/20 ms. A control group, 12 healthy volunteers, aged 20-59 years, were examined with the latter parameters. 128 accumulations, 1024 points, spectral bandwidth 1000 Hz were used. An unsuppressed water reference scan was included for chemical quantification. Automated quantification of absolute metabolite concentration was accomplished by LCModel. Results: MRI: In the group of patients and their descendants, one symptomless subject had a normal MRI (Fig.1), but all the other subjects had white matter changes (Fig.2a) at least under the motor cortex. MRS: In the pilot study, an abnormal peak was found in the scylloinositol region in all four subjects (two of those were symptomless) when the short echo time was used. In the total material, the abnormal peak (Fig.2b) was detected in 7/10 ADLD patients or their descendants. The differences between the controls and the study group examined using TR 6000 ms are presented in (Fig.3). Glucose and lactate were higher and NAA lower in the study group than in the controls. Conclusions: The abnormal peak in the scylloinositol region needs further studies. It may be an unknown metabolite simulating scylloinositol. The other spectroscopic changes were similar to those described in other leukodystrophies.

References: 1. Schwankhaus, Patronas, Dorwart, Elridge, Schlesinger, McFarland. [1998] Arch Neurol 45:1004-1008 2. Bergui, Bradac, Leombruni, Vauli, Quattrocolo. [1997] Neuroradiology 39:423-426 3. Raininko, Melberg, Weis. [2002] J Neurorad Suppl. 1S:220-221

Human Brain MRS: Paediatrics, Metabolic Disease, Focal Infections and Post Mortem

S25

49 1H NMR spectroscopy of mild (abortive) form of Canavan disease G. V. Vucurevic1, G. Kutschke2, J. Gawehn1, P. Stoeter1; 1Institute of Neuroradiology, University Clinic Mainz, Mainz, GERMANY, 2Children Hosptal, University Clinic Mainz, Mainz, GERMANY. Metabolism which involves cell-cell interaction (neuron-oligodendrocyte or neuron-astrocyte) cannot be obtained with spectroscopy of cell cultures. Spectroscopy in-vivo therefore can give add-on information in understanding of the role of metabolites such as Nacetyl aspartate, myo-inositol and lactate in neurochemistry observing pathological states in which these metabolite levels are altered. N-acetyl aspartate is recognized as a “neuronal” marker with the role that remains not completely understood. Normally, this amino acid gives the highest signal measured with 1H NMR spectroscopy of the brain. The only disorder in which there are observed increased values of this neuronal osmolyte is Canavan's disease. There are however more than 20 mutations of enzyme aspartoacylase (genetic information located in chromosome 17) found with rather similar clinical manifestations. Children with this disease are usually blind, paralyzed, and prone to seizures. Typical imaging appearance is massive spongioform leukodistrophy. Aspartoacylase (enzyme responsible for NAA catabolism) is located exclusively in oligodendrocytes (aspartate N-acetyltransferase, responsible for formation of NAA is present mainly in neurons). Recent research using cell cultures supports the role of catabolic products of NAA in oligodendrocytes in active myelin formation. The lack of this enzyme leads however to massive oligodendrocyte swelling. Patient: To our knowledge we present the case with the mildest form of Canavan disease which is not accompanied with spongiform leukodistrophy. Very mild neurologic manifestation and slight mental retardation were found in 17 year old girl with the family history of probably another case. Method: Localised single voxel 1H spectroscopy using STEAM (TE=20 ms) and PRESS (TE=135 ms) was performed in: basal ganglia, periventricular white matter and occipital gray matter. Results: Spectroscopies performed show normal metabolic concentrations (NAA 9 mmol) in gray matter, but spectroscopy from the region in white matter gives NAA elevated to 12 mmol. Except myo-inositol which was also significantly inreased both in white and in gray matter, other metabolites (Cholines, creatine and P-creatine) were not significantly different from normally found values in selected regions of the brain. Conclusion: Our findings are in agreement with the role of NAA in neuron-oligodendrocyte cellular interaction and efflux of NAA from gray to white matter.

50 MR proton spectroscopy of the brain in Wilson's disease W. Szeszkowski1, M. Golebiowski1, B. Tarnacka2; 12nd Dept. of Clinical Radiology, Central Hospital, Warsaw, POLAND, 2Neurology, Institute of Psychiatry and Neurology, Warsaw, POLAND. Purpose/Introduction: The goal of this study was to find out correlations of the brain metabolic changes in patients with according to different stages and subclinical encephalopathy in Wilson's disease, using MR proton spectroscopy. Subjects and Methods: Forty eight patients (31 women, 17 men; mean age 36 years, SD 11) with Wilson disease according to the diagnostic criteria of Sternlieb and Sheinberg (29 treated, 19 treated shorter than one year or untreated) were examined. According to the clinical picture the patients were divided into neurological (n=22), hepatic (n=14) and asymptomatic (n=12) groups. A control group consisted of 23 healthy volunteers (11 women, 12 men; mean age 29 years, SD 7). Brain MR images and 1H-MRS spectra were acquired on a 1.5-Tscanner. Single-voxel spectroscopy was performed in all 48 patients and 23 control subjects using PRESS sequences with parameters of 22/2000ms (TE/TR). Voxels (15x15x15 mm) were located in the center of the left and right globi pallidus and thalamus separately for each examination. The calculations were performed for the following metabolites: myo-Inositol(mI), Choline(Cho), Creatine(Cr), N-acetyl-aspartate(NAA), lipid(Lip), α-Glx(group from the amino acid CH at 3.7-3.9 ppm), and βγGlx(CH2-CH2 group at 2.1-2.5 ppm). These were expressed and analyzed as ratios to Cr. Results: In the entire patient population with WD, it was found a significant increase in Lip/Cr, Glx/Cr, Cho/Cr ratios and a decreased NAA/Cr level compared to the control group. That decreased concentrations of NAA were most explicit in the symptomatic group (neurological and hepatic symptom groups). The most pronounce increase of Lip, Glx and Choline levels for the neurological cases was observed. No statistically significant differences were found in mI/Cr ratios between WD patients and healthy volunteers, except for hepatic symptom group. The least changes of

S26

Human Brain MRS: Paediatrics, Metabolic Disease, Focal Infections and Post Mortem

metabolic concentration were detected in patients without clinical symptoms. Discussion/Conclusion: The decrease in NAA/Cr in our symptomatic population reflects neuronal loss. The decrease in mI though to be an early marker of portal - systematic encephalopathy was noted in patients with symptoms of liver disease. The increase of total lipids and Choline values especially in patients with neurological and asymptomatic forms of disease could reflect the changes caused by therapy or diet. The Glx increase in population of patients with WD is interpreted as indication of liver dysfunction. Very selective increase of Glx/Cr ratio in patients with neurological symptoms overthrow that hypothesis. These changes could also reflect the dysfunction of mitochondrial metabolism in WD.

51 Microbiological characterisation of focal brain infections by in vivo MR Spectroscopy U. Himmelreich1, N. Dorsch2, L. Gomes2, N. Jones3, G. Brown4, J. Taylor4, C. Mountford1, T. C. Sorrell2; 1Institute for Magnetic Resonance Research, University of Sydney, Sydney, AUSTRALIA, 2Westmead Hospital, University of Sydney, Sydney, AUSTRALIA, 3Department of Surgery, Royal Adelaide Hospital, Adelaide, AUSTRALIA, 4Department of Radiology, Royal Adelaide Hospital, Adelaide, AUSTRALIA. Introduction: Identification of microorganisms causing an infection is essential for optimal therapy. Current diagnostic methods are time consuming, invasive and require isolation of microorganisms from affected tissue or body fluids. Radiological methods used in routine diagnosis of cerebral mass lesions are unable to distinguish with certainty between different types of lesion (e.g. tumour, bacterial or fungal abscess). Glioblastoma multiforme remains part of the differential diagnosis on imaging. Distinction between glioblastoma and abscess was successful using Magnetic Resonance Spectroscopy (MRS). Aim and Methods: It was our aim to test the ability of in vivo MR Spectroscopy to distinguish between tumours and infective mass lesions as well as determine the microbial etiology of infective lesions. Single voxel MR spectra of infective cerebral mass lesions in humans were acquired on 1.5 Tesla MR scanners (Siemens Vision and GE Signa Horizon LX) at five Australian hospitals. MR spectra were acquired from 28 patients with focal brain infections and 35 patients with glioblastoma over five years. Correlation with microbiological findings and ex vivo MRS of pus/ biopsy samples was performed in all cases. Results and Discussion: MR spectra of all tumours showed an increased choline-to-creatine (3.2:3.0 ppm) resonance ratio (mean 2.5 ± 0.8). Necrosis was indicated by the presence of lipid resonances. MR spectra of patients with focal fungal or bacterial infections showed in all cases intensive lipid signals due to the presence of activated/apoptotic white blood cells. An increase in amino acid signal intensity was observed in 25% of focal infections. Intense signals of bacterial metabolites were present in abscesses. The metabolite profile correlated with bacteria identified in the pus (eg, lactate from Streptococcus milleri, acetate and succinate from anaerobic bacteria). MR spectra from patients with abscess caused by the same microorganisms were identical. MR signals found in vivo were confirmed by ex vivo MRS on biopsy and pus samples. It was possible to classify abscess according to groups of microorganisms (e.g., anaerobe, aerobe, fungal, tuberculoma, sterile)

Conclusion: Magnetic Resonance Spectroscopy is able to distinguish between tumours and infective mass lesions in vivo based on relative metabolite intensities. Classification of microorganisms causing infections is possible based on marker metabolites.

52 Statistical analysis of in situ 1H-NMR spectra of the brain leading to a quantitative determination of postmortem intervals (PMI) E. Scheurer1, M. Ith2, D. Dietrich3, R. Kreis2, J. Huesler3, R. Dirnhofer1, C. Boesch2; 1University of Bern, Institute of Forensic Medicine, Bern, SWITZERLAND, 2University of Bern, Department of Clinical Research, Bern, SWITZERLAND, 3University of Bern, Department of Mathematical Statistics, Bern, SWITZERLAND. Introduction: Estimation of the postmortem interval (PMI) is a crucial yet unsolved problem in forensic medicine, particularly in the later postmortem period. In situ 1H-MRS of decomposing brain in an animal model revealed unequivocal concentration changes of metabolites over at least 3 weeks postmortem [1]. The present work aims at a quantitative analysis of concentration changes of various metabolites in the brain by inverse parameterized functions, thus allowing calculation of PMI’s including statistical numbers. Materials and Methods: 8 sheep heads obtained during regular slaughtering and stored at constant temperature were examined in situ by 1H-MRS daily up to 18 days postmortem. 1H-MR Spectroscopy: 1.5 T (GE SIGNA), standard quadrature head coil, single voxel PRESS sequence (TR=3s, TE=20ms, water and outer volume suppression, eddy current correction, quantification via fully relaxed water signal, taking into account the GM/WM ratio in the voxel, fitting using LC Model. Statistical evaluation: For each measured metabolite an appropriate trend curve over time and its standard error was estimated. Time predictions and prediction intervals were derived by inversion of the model curves and their confidence bounds. Time predictions from different models for the same true time were combined by the generalized least squares method, weighted according to their variances [2]. Results: Parameterized model functions with the smallest number of parameters that still lead to an appropriate description of the time course were chosen. An example of concentration changes over time with applied model function and corresponding confidence intervals is shown in Fig.1. Propionate – thought to arise from bacterial metabolism - reveals an unequivocal function up to 400h. Fig.2 compares for every measurement time predictions combined from five metabolites (acetate, alanine, trimethylamine, butyrate, propionate) with true times after death. The correlation coefficients of the predicted versus true times are r=0.93 (all points) and r=0.97 (times <250h); bars indicate two standard deviations; deviation from straight line y=x occurs only at times >250h. Discussion and Conclusion: Analytical functions for the time course of ten metabolites in decomposing sheep brain tissue up to 400h postmortem have been defined. Predicted times combined from five metabolites correlate very well with true times postmortem up to 250h, while values above are systematically underestimated. This animal model shows the feasibility of a quantitative and objective determination of PMI’s. References: 1. Ith M et al. [2002] Magn. Reson. Med. 48:915-920 2. Sundberg R [1999] Scand. J. Statist. 26:161-207 Acknowledgements: Gebert-Rüf Foundation (GRS-069/99), Swiss National Foundation (31-59082.99)

Contrast Agents

S27

1.5T whole body scanner (Siemens, Vision) using a head coil. Informed consent was obtained before examinations. Several substances were tested in combination with different multislice and 3D imaging sequences. For instance, a 3D CISS sequence (12.3ms/5.9ms/70°) provided a high contrast using water (scan time: 4min, nom. res.: 0.6x0.6x1.0mm3). A contrast media (Magnevist enteral, Schering) was used in combination with a 3D FLASH sequence (11ms/2.5ms/60°; scan time: 2min8sec, nom. res. 0.6x0.6x0.4). Since the teeth appear as the most intense structure in the negative images, the data could be easily evaluated using standard 3D visualization tools such as isosurface rendering (AMIRA). Results: Best results were obtained using contrast media in combination with a FLASH sequence. Fig 1 compares a transversal slice of a data set acquired with and without contrast agent in the oral cavity and also shows the negative image. In fig 2 an isosurface reconstruction of the data is provided allowing an accurate spatial reproduction of the teeth and the jaw. Metallic and plastic inlays were not found to cause any imaging artifacts. Discussion: Substance aided dental-MR appears to be a valuable tool in dentistry and orthodontics and an interesting radiation free alternative to dental-CT. Since the data can be evaluated with standard dental software packages commonly used in dental-CT, the clinical application of dental-MR is straightforward. References: 1. Gahleitner A et al, Eur Radiol 2003, 13(2):366-76

Fig 1: Oral cavity native (a), with contrast agent (b) and after negation of the data (c)

Scientific Session Contrast Agents 3:30 pm - 5:10 pm

Plate

53 Substance aided dental-MR S. Olt, P. M. Jakob; Department of Physics, University of Wuerzburg, Wuerzburg, GERMANY. Introduction: In dentistry and orthodontics, dental-CT has greatly increased diagnostic possibilities [1]. The precise 3D visualization of the teeth and the jaw facilitates the pre-surgical planing of implants and therapeutics. However, due to the high radiation exposure, dental-CT can not be used as a routine clinical procedure. In this study we introduce dental-MR as a possible alternative method to image the different anatomical structures of dental systems in vivo with no need for ionizing radiation. As teeth are MR-invisible, our idea was to fill the oral cavity with a nonpoisonous substance which provides high signal (i.e. water, contrast media) so that the hard tissues become indirectly observable through contrast with the embedding medium. A beneficial side-effect is the minimization of the magnetic susceptibility effects at the mineralized tissue air interfaces in the oral cavity. Subjects/Methods: Eight healthy volunteers were examined on a

Fig 2: Isosurface reconstruction of two representative data sets

54 Preparation and characterization of MRI-detectable sterically, stabilized immunoliposomes W. J. M. Mulder1, L. van Bloois2, G. J. Strijkers1, G. Storm2, K. Nicolay1; 1Biomedical Engineering, Biomedical NMR, Eindhoven, NETHERLANDS, 2Pharmaceutics, Biopharmacy, Utrecht, NETHERLANDS. Introduction: A challenge in the screening of transgenic or diseased animal models is to develop new methods for non-invasively phenotyping and monitoring of disease progression1,2. Traditional

S28

Contrast Agents

immunohistochemical techniques can only be applied after fixation of the animal, followed by extensive biochemical procedures. This excludes the use of these techniques to follow progression of a disease in time and increases the amount of animals needed for statistical significance. In recent years, magnetic resonance imaging has emerged as the leading non-invasive in-vivo imaging modality. MRI, however, suffers from limited resolution and specificity, compared to traditional staining techniques. A way to overcome this problem is to use targeted contrast agents, which are directed to a molecular entity of interest (e.g. an endothelial cell surface receptor that is overexpressed as a consequence of the disease process). The low inherent sensitivity of MRI requires amplification of the signal. This can be achieved by the use of colloidal particles, which contain large amounts of Gd-DTPA. To that aim, we have developed contrast agents of sterically stabilized immunoliposomes, which contain Gd-based MRI contrast markers (figure 1). Materials and Methods: Liposomes containing DSPC, Cholesterol, Gd-DTPA-bis(sterylamide), PEG-DSPE, Mal-PEGDSPE were prepared by lipid film hydration. The liposomes were sized by extrusion and the size and size distribution were determined by dynamic laser light scattering. Antibodies were coupled to the distal ends of the Maleimide-PEG-chains as previously described3. T1-Relaxivity experiments were performed on a 6.3T MRI scanner. Results: The size of the liposomes was 103.7+/-0.9 nm before coupling of the IgG and 113.0+/-0.7nm after coupling of the IgG. The results of relaxivity measurements of free Gd-DTPA and the immunoliposomes containing Gd-DTPA-bis(sterylamide) are shown in figure 2. The introduction of PEG-lipids and the coupling of IgG had no effect on the relaxivity properties of the peparation. Discussion: This study shows the potential of these paramagnetic immunoliposomes as a long circulating contrast agent. The system is very flexible, can be sized down to 90 nm and is coated with PEG, which makes it applicable in vivo. The liposomes can be targeted due to the coupling of an antibody. In a similar manner peptides can also be coupled to the PEG chains4 and contribute to the specificity of this contrast agent. Inflammation, angiogenesis and apoptosis are examples of processes, which could possibly be detected with these liposomes in a non-invasive way with MRI. References: 1. R.Weissleder, Radiology,219,316(2001) 2. G.A.Johnson, JMRI,16,423(2002) 3. C.B.Hansen, BBA,1239,133(1995) 4. G.A.Koning, Int J Pharm,254,55(2003)

55 MRI of phagocytosis and receptor-mediated endocytosis using magnetic nanoparticles J. Wolter1, R. Mentlein2, U. Reus3, M. Heller4, C. Glüer1; 1University Clinics Schleswig-Holstein, Campus Kiel, Clinic for Diagnostic Radiology / Medical Physics, Kiel, GERMANY, 2University of Kiel, Institute for Anatomy, Kiel, GERMANY, 3GKSS Research Centre, Institute for Coastal Research / Physical and Chemical Analysis, Geesthacht, GERMANY, 4University Clinics Schleswig-Holstein, Campus Kiel, Clinic for Diagnostic Radiology, Kiel, GERMANY. Introduction: One major aim of molecular imaging is the visualisation of tumor-induced angiogenesis [1]. Angiogenic vessels, specifically their endothelial cells, express characteristic ligands such as vascular endothelial growth factor (VEGF) and its receptors (VEGFR) [2]. Some glioma cell lines also express VEGF and its receptor. The goal of our studies is the development of a site-directed contrast enhancement of tumor-induced angiogenic vessels in vivo. A Magnetic Resonance Imaging (MRI) contrast agent (iron oxide particle) is coupled to a VEGFR-antibody. Here we report results of the uptake of iron oxide particles by two different cell lines by receptor-mediated endocytosis as compared to non specific uptake based on phagocytosis. Materials and Methods: Two different cell types were used which express the VEGFR in different amounts: the endothelial cell line HUVEC (human umbilical vein endothelial cells) which highly express VEGFR-1 and VEGFR-2 and the human glioma cell line U118 which predominantly expresses VEGFR-2. VEGFR-1 and VEGFR-2 were coupled to two different types of ultra small dextran coated iron oxide paticles (70nm and 130nm NanomagParticle, Micromod GmbH, Rostock, Germany). The iron content in the cells was quantified by total reflection X-ray fluorescence (TXRF) analysis. The relaxation times of Nanomag-70nm and Nanomag-130nm were measured by MR-spectroscopy at 1.5T. Uptake was studied as a function of incubation time and iron concentration. Results: The following relaxation rates were observed: Nanomag70nm R1=11,4 1/(s*mM), R2=286,5 1/(s*mM); Nanomag-130nm R1=2,5 1/(s*mM), R2=188,3 1/(s*mM). The iron uptake by U118 cells by phagocytosis showed a maximum after 5h of incubation time, the HUVEC after 3h. These maxima were observed for both particles. In U118 the particle uptake of the Nanomag-70nm was dominated by phargocytosis, which was similar to the uptake by receptor-mediated endocytosis and about a hundred times higher than the Nanomag-130nm uptake. The HUVEC internalize more particles by receptor-mediated endocytosis especially by the VEGFR-2.

Contrast Agents

S29

U118 internalize approximately four times more particles compared to by receptor-mediated endocytosis by HUVEC. Discussion: Our results demonstrate that iron uptake of U118 by phagocytosis is much higher than the receptor-mediated endocytosis by HUVEC, which on the other hand has the advantage of higher specificity of the targeting process. Both mechanisms are promising for MR imaging of tumors and their angiogenic vessels. Our data allow an estimation of the required concentration of iron oxide particles required for visualization of neovascular growth by MRI. Literature: 1. Anderson SA [2000] Mag. Res. Med. 44:433-439 2. Yancopoulos G [2000] Nature 407:242-248

56 Study of the internalization of small particles of iron oxide by bone marrow cells S. Boutry, L. Vander Elst, R. N. Muller; Dpt Organic Chemistry/NMR Lab, University of Mons-Hainaut, Mons, BELGIUM. Purpose/Introduction: Nanoparticles of iron oxide have been recently used to magnetically label cells in vitro. Bone marrow contains not only haematopoietic precursors, but also stem cells. Such cells can therefore be implanted in animals and differentiate to repair injured tissues or organs, like brain (1). Magnetic labelling should allow for a MRI monitoring of cells migration after implantation. Our aim is to study the mechanisms of internalization of small particles of iron oxide (SPIO) and to quantify the internalization of these nanoparticles by bone marrow cells. Methods: Bone marrow cells were extracted from rat femurs and cultured during 12 to 15 days before an incubation of 48 hours with various concentrations of SPIO (2). Internalization of SPIO by bone marrow cells was measured by relaxometry at 60 MHz (Bruker Minispec mq60), T2-weighted MRI (Bruker Avance-200 imaging system, 4.7 T) and colorimetry after iron staining with Prussian blue on digested cells. For MRI and relaxometry, cells were resuspended in 2% gelatin. Phagocytosis and pinocytosis were inhibited with cytochalasin B and colchicine respectively (3). Results: Our results show that relaxometry, MRI and colorimetry allow for the quantification of internalized iron (Fig.1a). A decrease of the normalized R2 value of cells incubated with 100 µgFe/ml is observed in presence of colchicine and cytochalasin B (Fig.1b). Bone marrow contains phagocytic cells and non-phagocytic cells. The non-phagocytic cells are probably internalizing SPIO using pinocytic way.

Fig.1. Normalized R2 values (4.7 T) and internalized iron of cells incubated with different iron concentrations (a). Normalized R2 values (4.7 T) and internalized iron of cells incubated with 100 µgFe/ml; inhibitory effect of cytochalasin B and colchicine (b). Conclusions: It is possible to evaluate the iron quantity internalized by cultured bone marrow cells. These measurements can be applied to the study of the internalization of specifically targeted nanoparticles of iron oxide by other types of cells. These preliminary data seem to show that pinocytosis and phagocytosis are co-existing in bone marrow. Both ways of internalization can therefore be used by bone marrow cells to take up SPIO. References: 1. Zhao L.R. et al. [2002], Exp. Neurol., 174, 11-20 2. Herynek V. et al. [2002], MAGMA, 15 Suppl 1, 16 (19th Annual Meeting European Society for Magnetic Resonance in Medicine and Biology, Cannes, France) 3. Fleige G. et al. [2002], Invest. Radiol., 37, 9, 482-488

57 Dynamic contrast-enhanced MRI of tumor angiogenesis with different generations of Gd-DTPA terminated dendrimers Q. G. de Lussanet1, S. Langereis2,3, R. G. H. Beets-Tan1, M. H. P. van Genderen2,3, A. W. Griffioen4, J. Smits5,3, E. W. Meijer2,3, J. M. H. van Engelshoven1,3, W. H. Backes1,3; 1Radiology, University Hospital Maastricht, Maastricht, NETHERLANDS, 2Laboratory of Macromolecular and Organic Chemistry, Eindhoven University of Technology, Eindhoven, NETHERLANDS, 3Biomedical Engineering, Eindhoven University of Technology, Eindhoven, NETHERLANDS, 4Angiogenesis Laboratory of Pathology, University Hospital Maastricht, Maastricht, NETHERLANDS, 5Pharmacology, University Hospital Maastricht, Maastricht, NETHERLANDS. Angiogenesis, new blood vessel formation, is essential for tumor growth. Non-invasive techniques, such as dynamic contrast-enhanced magnetic resonance imaging (DC MRI) with Gd-DTPA can be used to calculate parameters that reflect angiogenic activity in the tumor [1]. Poly(propylene imine) dendrimers are monodisperse, highly branched macromolecules, which are synthesized in a stepwise fashion. The multivalent nature of dendrimers allows conjugation of a discrete number of Gd-DTPA moieties to the functional groups at the periphery of the dendrimer [2]. This study compared the use of different generations of Gd-DTPA terminated poly(propylene imine) dendrimers, i.e. different sizes of contrast agent, for DC MRI of tumor angiogenesis. Generations 0, 1, 3, and 5 of poly(propylene imine) dendrimers were modified with gadolinium chelating moieties (DTPA) and complexated with gadolinium, respectively. Longitudinal relaxivi-

S30

Contrast Agents

ties (r1) were determined in vitro at 1.5 T and 21°C. Tumor (LS 174T) bearing mice (Swiss nu/nu) were used for DC MRI measurements. MRI was performed using a surface coil (5 cm diameter) and 1.5 T MRI system [1]. Either agent G0, G1, G3 or G5 was injected in the tail vein (0.03 mmol Gd / kg in 3 mM citric acid buffer (5 microl). Applying a 2-compartment model yielded a microvessel permeability surface area product (KPS, ml min-1 100 cm-3 tissue) for each voxel in a tumor [1]. Our results show that lower generations dendritic contrast agents result in higher measured KPS values compared with higher generations contrast agents (Spearman correlation, p = 0.04) [Figure 1]. These differences may be related to the specific kinetic physiologic and/or magnetic properties of each agent [2]. DC MRI showed faster clearance from circulation through the kidneys for G0 and G1 than for G3 and G5, and G5 showed a substantial residence time in both liver and spleen. In vitro measurements showed increasing longitudinal relaxivity (r1) values for increasing generation of dendrimers, ranging from G0=8.6 to G5=20.0 mM-1s-1. The comparison of low and high molecular weight dendritic contrast agents shows that each of the generation 0, 1, 3 and 5 can be used for mapping tumor angiogenesis with DC MRI, and that there is a difference between the measured KPS values of the lower generations and the higher generations dendritic contrast agents. Additional studies should further explore possible differences in kinetic physiologic and magnetic properties of the agents. 1. de Lussanet, QG et al. [2003] Radiology in press 2. Kobayashi, H et al. [2003] Bioconjugate.Chem 14:388-394.

58 Manganese ions released from manganese DipyridoxylDiphosphate provide very strong intracellular T1 relaxation in rat myocardium W. Nordhøy1, H. W. Anthonsen2, M. Bruvold1, J. Krane2, H. Brurok1, P. Jynge1; 1Department of Circulation and Imaging, Faculty of Medicine, Trondheim, NORWAY, 2Department of Chemistry, Faculty of Natural Sciences, Trondheim, NORWAY. Introduction: Teslascan™ containing manganese (Mn) dipyridoxyl-diphosphate (MnDPDP) is a liver contrast medium that works by the release of paramagnetic Mn2+ ions after intravenous infusion. These ions are also taken up into cardiomyocytes where they are retained for hours. In the present study we wanted to obtain knowledge about the efficacy of Mn2+ ions as intracellular (ic) contrast agents for the myocardium. Subjects and Methods: Relaxation parameters, T1 and T2, and Mn content were measured in right plus left ventricular tissue excised from isolated perfused rat hearts (N = 24). In the experiments

5 min wash-in with MnDPDP 0, 30, 100, 300 and 1000 µM (n = 46) were followed by 15 min wash-out to remove extracellular (ec) Mn2+. Results and Conclusions: Physiological monitoring revealed no cardiodepressive action during wash-in with MnDPDP below 1000 µM. MnDPDP 100, 300 and 1000 µM raised tissue Mn content to 1.6, 2.4 and 5.2 times control. Inversion recovery analysis at 20 MHz revealed two components for longitudinal relaxation: a short T1-1and a longer T1-2. Mean values for control and Mn-treated hearts ranged from 633 to 233 ms for T1-1 and from 2501 to 985 ms for T1-2 and the intensities were about 65 % and 35 % for the short and the long component, respectively. Tissue Mn content correlated well with myocardial relaxation rates for R1-1 (r = 0.921) and less well for R1-2 (r = 0.691) (see Figure).

Two-site water exchange analysis documented a slow water exchange regime in the present excised cardiac tissue and, accordingly, no differences between apparent and intrinsic relaxation parameters. A most important finding was that ic relaxivity induced by Mn2+ ions in the ic water compartment (r1-1) was as high as 58 (s mM)-1. This level of relaxivity is about 7 times and 31 times higher than measured with respectively MnCl2 and MnDPDP dissolved in pure water. We conclude that strong interactions between ic Mn2+and macromolecules (proteins) with greatly reduced rotational movements and maintained inner sphere water access make Mn2+-ion-releasing intracellular contrast agents surprisingly effective for T1 weighted imaging. Accordingly, MnDPDP, a slow-to-intermediate Mn2+ ion releaser, should be a promising agent for cardiac MRI.

59 Chemical shift angiography (CSA) with dysprosium-DOTA S. Miraux, E. Thiaudière, P. Canioni, J. Franconi; Rmsb, CNRS/Universite Bordeaux2, Bordeaux, FRANCE. Introduction: Nowadays Magnetic Resonance Angiography is realised almost exclusively with a Gadolinium injection. However, contrast between vessels and stationary tissues is limited by the Gadolinium dose. Thus, small vessels might be hardly visible with this technique. The aim of the present work was to develop a new MRA technique based on the use of a chemical shift contrast agent allowing to obtain new kind of contrast not based upon relaxivity parameters. Dysprosium is well known to modify the resonance frequency of water molecules situated at its proximity. In the millimolar range, the induced chemical shift is directly proportional to Dysprosium concentration. For example, Dy-DOTA at a concentration of 5mM shift water resonance frequency of 1 ppm.

Functional MRI Materials and Methods: Phantoms were made of vials containing several Dysprosium concentration (0mM, 5mM, 10mM, 25mM and 50mM) corresponding to 0 to10 ppm chemical shifts with respect to pure water. Experiments were carried out on a 4.7 Tesla Bruker Biospec 47/50 (Bruker, Germany). The system was equipped with a 12 cm gradient system of 193 mT/m maximum strength. In order to suppress non-shifted water signal, a binomial pulse (1-3-3-1) follow with a spoiling gradient is introduced before a 3D gradient echo sequence (Matrix size = 128×128×32 ; FOV = 4×4×3.5 ; TR/TE = 12/2 ms). In vivo experiments were carried out on male Wistar rats (250 g) injected with 1.2 mmol/kg Dy-DOTA (Guerbet, France). Results: Binomial pulses were adjusted on Dysprosium phantoms in order to both eliminate the pure water signal and select the proper chemical shifted frequency. This was achieved by changing the delay between the pulses. The remaining water signal was estimated at about 0.1%. Binomial delay could even be adjusted to visualized more than one Dy-DOTA concentrations. High dose Dysprosium injection in living rats allowed a clear visualisation of the carotid arteries by selecting a chemical shift of 300 Hz (1.5 ppm). However the contrast was effective only during a single volume acquisition, probably because of Dy-DOTA dilution by circulation. Discussion: The first results achieved on phantom and in vivo show the potential interest of this new technique. With the suppression of non-shifted water, a better contrast between vessels and tissues was obtain than from Gd-DOTA injection at 4.7 T. Further developments are required in order to routinely use this technique for imaging very small vessels.

60 2D steady state free precession (SSFP) magnetic resonance angiography (MRA): A comparison with contrast-enhanced 3D MRA in patients with aortic coarctation P. A. Wielopolski1, J. van den Berg, M.D.2, R. J. van Geuns, M.D., Ph.D3, P. M. T. Pattynama, M.D, Ph.D.1, W. A. Helbing, M.D, Ph.D.2; 1Radiology, Erasmus Medical Center, Rotterdam, NETHERLANDS, 2Children Cardiology, Erasmus Medical Center, Rotterdam, NETHERLANDS, 3Cardiology-Thorax Centrum, Erasmus Medical Center, Rotterdam, NETHERLANDS. Introduction/Purpose: Steady state free precession (SSFP) imaging has generated renewed interest with the introduction of faster and stronger gradient subsystems. High signal-to-noise (SNR) and contrast-to-noise (CNR) ratios can now be achieved with SSFP readouts with ultrashort repeat times that have proven excellent for MR angiography (MRA) without contrast agents. The purpose of this work is to evaluate the diagnostic value of ECG-gated singleshot 2D SSFP MRA using 3D contrast-enhanced (CE) MRA as standard of reference, in patients with aortic coarctations. Methods and Materials: We scanned 10 patients (2 children, 8 adults) using ECG-gated (mid-late diastole) fat suppressed, noncontrast enhanced sequential 2D SSFP MRA (TR=3.3-3.6 ms/TE=1.5-1.7 ms/α=100-140o/SL=3 mm, spacing=1 mm). In 9 out of 10 patients CE MRA was performed for comparison (TR=4.3 ms/TE=1.4 ms/ α=30o/SL=2.4 mm/spacing=1.2 mm/ single dose Gd-DTPA). Sagittal scans were performed. For SSFP MRA, additional planes were obtained. Multi-planar reformations were used for data inspection. A General Electric 1.5 T MRI was used with a torso array coil. All data were acquired in 1-2 breathholds at end-expiration. For SSFP MRA 15–22 slices per breathhold (1 slice/ heartbeat) were selected according to heartrate.

S31

Results: The aortic arch could be visualized in all patients. CE MRA showed moderate ghosting from the thoracic aorta in all patients. 2D SSFP MRA was respiratory motion free, however, inconsistency between breath-holds could diminish the quality of a 3D reconstruction. High post-stenotic velocities provided limited to severe dephasing of blood signal in SSFP MRA, and to a lesser extent in CE MRA. On the other hand, contrast timing for CE MRA was not always optimal. For SSFP MRA, inhomogeneous blood signal was graded none-mild in 6, moderate in 3 and severe in 1, while dephasing was graded none-mild in 7, moderate in 2 and severe in 1. Signal loss could be reduced by optimizing trigger delay in SSFP MRA (2 cases). Inhomogeneity/contrast enhancement/dephasing in CE MRA provided similar data: none-mild in 8 and moderate in 1. Diagnostic value of SSFS MRA was good in 8, moderate in 1 and insufficient in 1 case. Using CE MRA diagnostic value was good in 8 and insufficient in 1. SNRs and CNRs were usually higher for CE MRA. Discussion/Conclusion: The diagnostic value of SSFP MRA using CE MRA as gold standard is similar. Patient studies where contrast is not strictly required for the evaluation of the thoracic aorta, SSFP MRA can suffice for an appropriate diagnosis.

Clinical Focus Session Functional MRI 5:30 pm - 6:30 pm

Willem Burger

61 A functional magnetic resonance imaging study of cognitive estimation test J. Tintera1, J. Horacek2, M. Preiss2; 1Department of Radiology, IKEM, Prague, CZECH REPUBLIC, 2Prague Psychiatric Centre, Charles University, 3rd Medical Faculty, Prague, CZECH REPUBLIC. Introduction: Cognitive Estimation Test (CET) is the diagnostic and research tool for quantification the ability to provide accurate cognitive estimates. The cognitive estimation is supposed to be responsible on the prefrontal cortex activity and cognitive estimation declines in patients with prefrontal lesions. The general aim of our study was to evaluate the brain regions responsible for cognitive estimation by the use of functional MR imaging (fMRI). Subjects and Methods: fMRI measurements were done on Siemens Vision 1.5T with GE EPI sequence (TE=54ms, TR=7s, α=90°). Block design was used: 27 slices (THK=4mm), 64 images, 8 images per each period of rest and stimulation. Under the experimental condition, subjects were instructed to overtly estimate the answer to visually presented questions presented in blocks. The active (estimation) blocks for the cognitive estimation consists of 32 CET tasks and the rest (not-estimation) blocs of the trivial questions not requiring any estimation. Evaluation was done in SPM 99: Realignment, smoothing and normalization, square-box function (for both 1st (subject) level and 2nd (group) level, uncorrected p=0.0001 was used). Ten right-handed volunteers (7 women and 3 men, average age 24.8 years, 21-33 years) participated in the study. Results: In the group of ten healthy volunteers we used functional magnetic resonance imaging to detect the brain systems involved by cognitive estimation. We have confirmed that the Cognitive Estimation Test changes the fMRI signal in frontal, parietal and occipital cortex. The BOLD signal was increased in active (estima-

S32

Functional MRI

tion) condition mostly in dorsolateral prefrontal cortex (left Inferior Frontal Gyrus and left Middle Frontal Gyrus BA 9) and occipitoparietal cortex (fig. 1a). On the other hand, in active (estimation) condition, the BOLD signal has decreased in the medio-prefrontal part (left Medial Frontal Gyrus BA 10, right Precentral Gyrus BA 62, right Anterior Cingulate BA 32 and left Mediofrontal white matter), Caudate and Precentral Gyrus (fig. 1b).

Fig.1. a) Areas of activation and b) “deactivation” during CET. Conclusion: Our data indicate that the activation of medioprefrontal and deactivation of left dorsolateral prefrontal cortex and occipito-parietal lobules are the neurobiological substrates of cognitive estimation. Study was supported by grant MSMT CR LN00B122.

62 FMRI-study of dual task working memory performance M. Ries1, S. MacPherson1, G. Waiter1, S. Della Sala1, R. Logie1, S. Cappa2; 1Cognitive Neuroimaging Research Unit, School of Psychology, Mathematics and Computing, Aberdeen, UNITED KINGDOM, 2Department of Psychology, S. Raffaele University, Milano, ITALY. Introduction: A fundamental question about the nature of cognitive control is whether performing two tasks simultaniously is associated with additional brain areas for coordination or with more widespread activity of areas associated with performing each task separately1. This study compares the activity ot two working memory tasks, consisting of a visual pattern memory task and an auditory verbal letter task during a retention phase in which both tasks are in concurence, with the corresponding single task activity2. Methods: The dual task paradigme (Fig 1) was carried out by 32 participants twice (six blocks per session, single task order swaped, stimulus order randomized). Good performers only (70% success rate) entered data analysis of the retention phase (n=21/18). In addition all participants compeleted the single task paradigms for comparison. Scan parameters: EPI on a 1.5T GE-Signa with TR =2.5s, TE =35ms, 128x96 matrix, 24 slices, 2x2x5mm resolution. Data analysis: Realignment, smoothing (8mm), normalisation and t-Tests were carried out with SPM993 with a final random effect analysis on the two groups (maxima at p<0.005 uncorrected are reported).

Figure 1: The paradigme allows the dual task activity of the memory phase to be contrasted against the single task activation. Results: While the single letter task showed activity in the anteriour cingulate(AC,Broadman Area 32), the medial frontal gyrus (MFG, BA6), the inferiour frontal gyrus (IFG, BA46/47) and the superior middle frontal gyrus (SMFG, BA46), the single pattern task showed only activity in the IFG(BA47). As shown in Fig 2, the dual memory task showed increased activity in the IFG in BA47, new activity in the IFG in BA45 and the superior frontal gyrus (SFG, BA8) and modified activity in the AC and the SMFG(BA9).

Figure 2: Contrast between dual task and single task experiments. Conclusions: The cognitive demand of performing two working memory tasks concurrently with the recruitment of additional cortical areas in the IFG (BA45,47) and, to a lesser extent, by modifications in the activity of the "classical" working memory areas in the SMFG (BA9,46) as suggested in previous studies1,4. References: 1. D'Esposito M. [1995] Nature, 378:279-281. 2. Coccini G. [2002] Memory and Cognition, 30: 1086-1095. 3. Friston. Human Brain Function, 25-41 (Academic Press, London). 4. Dreher JC. [2003] Cerebral Cortex, 13:329-339.

Functional MRI 63 Compensatory cortical activation observed by fMRI during a high cognitive task at the earliest stage of MS S. Confort-Gouny1, B. Audoin2, D. Ibarrola1, J. Ranjeva1, I. Malikova2, A. Ali-Cherif2, J. Pelletier2, P. J. Cozzone1; 1Crmbm umr cnrs 6612, Faculte de Medecine, Marseille, FRANCE, 2Departement de Neurologie, CHU Timone, Marseille, FRANCE. Purpose: To determine whether cortical reorganization occurs during high cognitive processes at the earliest stage of MS. Subjects and Methods: we performed an fMRI experiment using the Paced Auditory Serial Addition Test (PASAT) as paradigm in a population of ten patients with clinically isolated syndrome suggestive of multiple sclerosis (CISSMS) fulfilling at inclusion at least the dissemination in space criteria according to McDonald (dissemination in space demonstrated by MRI, or positive CSF plus two or more MRI detected lesions consistent with MS). Eight patients had a diagnosis of MS and two had a diagnosis of 'possible MS' according to McDonald’s criteria when comparing the first MRI performed after relapse and conventional MRI performed at the time of the fMRI exploration at least 3 months after onset. At the time of the fMRI exploration, mean disease duration was 6.8 + 3.3 months. We compared these results to those obtained in a group of ten education-, age- and sex- matched healthy controls. Subjects were explored on a 1.5 T MRI system using single-shot gradientecho EPI sequence. Performances of the two groups during PASAT recorded inside the MR scanner were not different. Results: Statistical assessment of brain activation was based on the random effect analysis (between-group analysis two-sample t-test p<0.005 confirmed by individual analyses performed in the surviving regions p<0.05 Mann Whitney U-test). Compared to controls, patients showed significant greater activation in the right frontopolar cortex, the bilateral dorsolateral prefrontal cortices and the right cerebellum. Healthy volunteers did not show greater activation compared to CISSMS patients. Discussion and Conclusion: The present study argues in favor of the existence of compensatory cortical activations in CISSMS patients performing PASAT mainly located in regions involved in executive processing. It also suggests that fMRI can evidence the active processes of neuroplasticity contributing to mask clinical cognitive expression of brain pathology at the earliest stage of MS.

64 Modulatory effects of stimulus intensity and attention on the BOLD-response elicited by mechanoreceptor stimulation A. M. Sterr1, A. Zaman2, W. Bimson2, C. Hayward1, N. Roberts2; 1Center for Cognitive Neuroscience, University of Liverpool, Liverpool, UNITED KINGDOM, 2Magnetic Resonance and Imaging Analysis Research Center, University of Liverpool, Liverpool, UNITED KINGDOM. Purpose: The somatosensory system has been widely used as a ‘model’ modality to study adult human brain plasticity, mainly by means of mapping experiments. However, the neural correlates of tactile information processing, in particular for ecologically valid activation of mechanoreceptors, remains to be characterized. Here, we report preliminary results of a study that aimed to answer two questions. 1. How is stimulus intensity reflected in the hemodynamic response, and 2. how is this response modulated by attention.

S33

Materials: Tactile stimuli were presented to the midvolvar surface of the middle phalanx of the right index finger. An electronicallycontrolled device was used to deliver prodding events that activated the mechanoreceptors in their physiological range. Four experimental conditions (2 x 2 – design), varying stimulus intensity (high/low) and attention (attend/ignore tactile events), and an active control condition (ACC) were employed. All experimental runs used an oddball paradigm (standards: 813 ms SOA, deviants: 640 ms SOA; 0-2 deviants/ run). In the attend condition, subjects were asked to silently count the deviants; in the ignore condition a mental arithmetic task had to be performed. The arithmetic task but no tactile stimulation comprised the ACC. Participants were blindfolded during the experiment. Imaging data was collected with a 1.5Tesla GE (LX/Nvi NeuroOptimised) MRI system. BOLD contrast was obtained from T2*weighted EPI encompassing the whole cerebrum [TR 3s; TE 40ms; ^ 90?; FOV 24 cm, slice thickness 5 mm, 22 slices], and superimposed on a T1-weighted 3D image for anatomical localisation. 100 EPI volumes per condition were acquired in a 10-cycle boxcar design with 15s ACTIVE/15s BASELINE. Conditions were counterbalanced across participants. SPM99 was used to conduct single subject (p<0.001) and group conjunction (p<0.05) analysis. Results: Contralateral S1 activation was most prominent in all conditions. In addition, significant activation was found in ipsilateral S1, SMA, right IP and right STG in the attend condition, and ipsilateral S1 and bilateral IFG when stimuli were ignored. With attention, the activated volume and the BOLD response increased in S1, while higher stimulus intensity caused an increase in signal strength only. Conclusion: Attention and stimulus intensity differentially affect the hemodynamic response in S1. Higher stimulus intensities are accompanied by a stronger BOLD response, which is in line with findings in the motor system. With attention, a greater region in S1 is active. This implies that future mapping studies need to control for attention in order to identify ‘true’ changes in homuncular organisation.

65 Bilateral cortical representation of the trunkmidline in primary somatic sensory areas of the human brain G. Polonara1, M. Fabri2, T. Manzoni2, U. Salvolini1; 1Department of Neuroradiology, Università Politecnica delle Marche, Ancona, ITALY, 2Institute of Human Physiology, Università Politecnica delle Marche, Ancona, ITALY. Purpose: This research aims at studying the cortical representation of the trunk zone in the area SI of human brain, to establishing whether in man, as in non-human primates, the cutaneous regions close to the trunk midline are represented in the primary somatic sensory cortex of both hemispheres. Subjects and Methods: eight volunteers and four callosotomized patients gave their informed consent to participate in the study, approved by the Ethical Committee of the University of Ancona. One patient has complete callosal resection, three have partial resection sparing the splenium and probably the posterior part of the callosal trunk. The study has been carried out using a General Electric Signa LX NV/i magnet (1.5 Tesla, gradients 50mT/m). The images of 10 contiguous 5-mm-thick encephalic sections parallel to the axial plane have been acquired using an echo planar sequence. Unilateral tactile stimulation was applied by rubbing cutaneous ventral trunk

S34

Breast: Imaging, Perfusion and Spectroscopy

regions close to the midline (approximately 20 x 10 cm in width) with a soft cotton pad (frequency 1 Hz). The paradigm lasted 5 minutes and consisted of ten 30-sec alternating periods of rest and stimulation. Results: In control subjects cortical activation foci to the unilateral stimulation were found in the post-central gyrus (PCG) of both hemispheres, slightly medial to or just behind the omega-shaped region of the central sulcus (CS), and in the anterior bank of the post-central sulcus (PCS). These regions correspond to the trunk representation zones of areas 1 and 2-5, respectively. In the complete callosotomised and in one of the three partial callosotomised patients activation foci were present only in the contra lateral PCG, at variance of that observed in subjects with intact corpus callosum. In the other two patients with partial resection activation foci were present in the PCG of both hemispheres, as observed in control subjects. Conclusion: Present data confirm that in man, as in non-human primates, the cutaneous regions adjacent to trunk midline are represented bilaterally in SI, and indicate that the ipsilateral activation requires the integrity of the posterior part of the corpus callosum.

66 Functional MRI and diffusion tensor imaging in a patient with central pain M. L. Seghier1, F. Lazeyras1, P. Vuilleumier2, A. Carota3; 1Dept. Radiology, University Hospital of Geneva, Geneva, SWITZERLAND, 2Laboratory of Neurology and Imaging of Cognition, University Medical Center, Geneva, SWITZERLAND, 3Clinique de Reeducation, University Hospital of Geneva, Geneva, SWITZERLAND. Introduction: PET and fMRI studies in healthy subjects suggest that pain processing may involve the insula, cingulate gyrus, thalamus and parietal areas. By contrast, PostStroke Central Pain (PSCP) has been less often investigated by functional studies. The link between the lesion and activity in the network processing emotional and cognitive aspects of pain is not yet clearly established. Here, we present a single-case study of a patient with PSCP in whom fMRI and DTI techniques allow more detailed anatomoclinical correlations. Subjects and Methods: A 65 years old man showed clinical features of PSCP (hypoesthesia, hyperpathia and cold allodynia) of the left body, 3 years after an hemorrhagic stroke involving the corona radiata adjacent to the postero-lateral thalamus. Pain was intense, selectively triggered by cold stimuli, and only partially attenuated by pharmacological therapy. During fMRI we used a block paradigm consisting of three conditions in pseudo-random order (each of 24sec duration), in which stimuli with three different temperatures (warm "W", cold "C", and freezing "F") were applied on either the right or left hands, separately stimulated. Functional data consisted of 29 contiguous 5-mm axial slices, collected with a 1.5T EPI sequence (TR=3sec). For the DTI acquisition, SE-EPI sequence (TE=65ms, b-factor=1000s/mm2, 6 non-collinear directions) was used. Functional data were analysed using cross-correlation method. Diffusion data, after geometric distortion correction, were decomposed in eigenvectors for delineation of fibre tracts. Results: Three main pattern of responses were detected: 1) the medial superior frontal gyrus (BA 6) and primary somatosensory cortex (SI) showed similar signal changes during the three conditions W, C and F; 2) increasing intensity responses when stimulus temperature was increasingly cold were observed in the controlateral

hemisphere in the posterior insula and SII; 3) robust responses were found specifically during the F condition (when a sensation of pain was evoked by the left hand stimulation) in the anterior cingulate cortex (BA 24), associative cortex of the parietal lobe (BA 7) and putamen. DTI maps show a loss of fibres in the left corona radiata adjacent to the lesion and a reduction of fibre density in the posterior left insular region facing the residual hemorrhagic cavity.

Conclusion: We suggest that, for this patient, the anatomo-functional substrate of the cold hyperpathia is due to the disconnection between the lateral thalamus and the posterior insula and the resulting activation of the parietal associative areas (intensity pain processing) and anterior cingulate cortex (emotional pain processing).

Clinical Focus Session Breast: Imaging, Perfusion and Spectroscopy 5:30 pm - 6:30 pm

Jurriaanse

67 Role of MR mammography in detection of intraductal spread in carcinomas with extensive intraductal component M. L. A. Van Goethem1, K. Schelfout2, E. Kersschot3, C. Colpaert2, J. Weyler4, I. Verslegers1, I. Biltjes1, I. Eerens1, A. M. A. De Schepper1; 1Radiology, University Hospital Antwerp, Edegem, BELGIUM, 2Pathology, University Hospital Antwerp, Edegem, BELGIUM, 3Radiology, OLV Hospital Aalst, Aalst, BELGIUM, 4Epidemiology and Social Medicine, University Antwerp, Edegem, BELGIUM. Purpose: To determine the possibility of MR mammography to detect extensive intraductal component accompanying invasive ductal carcinoma (EIC + tumors). Introduction: Locoregional staging of breast carcinoma is necessary to select patients for breast conserving therapy and include accurate assessment of tumor size, multifocality and the presence of an extensive intraductal component. EIC + is responsible for higher recurrence rate after breast conservative surgery. Material and Methods: In a prospective study 231 consecutive women with a suspicious lesion diagnosed on clinical examination, mammography and/or ultrasound underwent preoperative MR mammography and 296 malignant lesions were detected. Morphology and extent of lesions on MR mammography was correlated with findings on pathological examination. Results: Our study population consisted of 195 invasive ductal carcinomas, 50 of them were EIC positive (25,6%) (37grade III, 4 grade II, 8 grade I).

Breast: Imaging, Perfusion and Spectroscopy Mammography predicted intraductal carcinoma in 48,5% (62,5% grade III, 100% grade II, 16,6% grade I), and MR in 68% (71% grade III, 25% grade II, 75% grade I). On MR mammography 22 of them presented as an area of enhancement with ductal or linear pattern, 14 as long spicules, 5 as a regional enhancing area, 8 had small nodules adjacent to a mass (15 had a combination of enhancing structures accompanying the mass). In 11 tumors we studied enhancement kinetics of the surrounding area. Nine of them showed more than 100% intensity increase, 6 showed continuous increase in time, 2 had a wash-out and 3 showed a plateau. Discussion: Our findings are in concordance with previous studies. Mumtaz et al reported on 53 cancers, of which 19 were EIC+. Sensitivity of MR mammography (76%) for demonstration of DCIS was higher than of mammography (52%). If EIC + tumors were seen MR showed an adjacent area of diffuse or linear enhancement around the mass. Satake reported on a sensitivity for ductal spread of 93% patients on MR mammography on 25 patients. Conclusion: In our study MR mammography had the highest sensitivity to predict intraductal spread. Moreover the prospective character of the study strengthens the results.

S35

the inclusion of PK parameters increased diagnostic power significantly, over-and-above that obtainable with pre-MRI data only (unequivocal diagnosis being possible in 53% of cases), the inclusion of textural and morphometric parameters did not. Discussion/Conclusion: PK and LRAM analysis of DCE-MRI was able to provide an accurate, unequivocal diagnosis in 53% of C3/C4 cases which could permit rapid, semi-automated diagnosis and a useful role in second reading. Despite displaying some diagnostic power in isolation, textural and morphometric parameters did not provide added value in the evaluation of C3/C4 cases by MRI. 1. D.J. Manton Proc. ISMRM 7 abstr. 1079 (1999); 2. P. Gibbs Proc. ISMRM 10 abstr. 1910 (2002); 3. P. Tofts Magn. Reson. Med. 17 357 (1991); 4. R.M. Haralick Proc. IEEE 67 786 (1979)

68 Equivocal breast cytology: Does quantitative textural and morphometric analysis of contrast-enhanced MRI have added value? D. J. Manton, P. Gibbs, P. Kneeshaw, L. W. Turnbull; YCR Centre for Magnetic Resonance Investigations, University of Hull, Hull, UNITED KINGDOM. Purpose/Introduction: In breast cancer patients with equivocal cytology (C3/C4), dynamic contrast-enhanced (DCE-) MRI can provide accurate, unequivocal diagnosis in 50% of cases; potentially avoiding the physical and psychological trauma associated with additional surgical biopsy [1]. Cases where DCE-MRI could not provide definite diagnosis included highly vascular benign lesions (which cannot be differentiated from cancers by contrast uptake dynamics alone). Textural analysis of static contrast-enhanced MRI of breast tumours also has diagnostic power [2] and a retrospective study was carried out investigating whether this technique could improve diagnostic accuracy in the C3/C4 patient sub-set. Subjects/Methods: 38 patients (14 with histologically proven malignancies) were imaged subsequent to a C3/C4 result. DCE-MRI involved bolus injection of Gd-DTPA and T1-weighted FSPGR imaging (1.5 T IGE Signa, 11.6 s temporal resolution, 7 minute scan). T1-weighted 3D, post-contrast, fat-suppressed images were also acquired (24 cm FOV, 512 x 256 matrix). A three compartment pharmacokinetic (PK) model was used to analyse DCE-MRI [3]. 3D images were decimated to 32 grey levels then analysed using co-occurrence matrices [2, 4]. Lesion cross-sectional area and circularity were also quantified. Parameters were combined, to obtain a synergistic increase in diagnostic efficacy, using logistic regression analysis modelling (LRAM). Results: All pre-MRI parameters (cytology grade, mammography grade, patient age) demonstrated substantial overlap of benign and malignant data. With some MRI parameters, however, data were sufficiently well separated to permit accurate, unequivocal diagnosis (i.e. with 100% positive and/or 100% negative predictive value). This could be achieved in 16% of cases for MRI grade (expert radiologist), 42% for PK exchange rate (see graph), 21% for lesion area and 16% for lesion texture. LRAM indicated that whilst

69 Dynamic MR imaging of the breast: Analysis of diagnostic criteria using artificial neural networks B. K. Szabo, M. Kristoffersen Wiberg, B. Bone, P. Aspelin; Division of Radiology, Center for Surgical Sciences, Karolinska Institutet, Huddinge University Hospital, Stockholm, SWEDEN. Purpose: To assess the discriminative ability of different established diagnostic criteria for MRI of the breast, and to determine their relative relevance using artificial neural networks. Subjects and Methods: A total of 89 women with 105 histopathologically verified breast lesions were included in this study. Patients were examined by a 1,5 T system using a dedicated double breast coil. A dynamic examination with one precontrast and seven postcontrast series was performed, using a T1-weighted three-dimensional fast low-angle shot (3D FLASH) sequence. Architectural features as well as parameters derived from time-signal intensity curves were analyzed. Feed-forward back-propagation neural network models were constructed using the NevProp 4R1 software. The subjects were randomly divided into two parts: a training set of 59 lesions and a verification set of 46 lesions. Training set was used for learning, and the performance of each model was evaluated on the verification set by measuring the area under the ROC curve (C-index). Different models were built to test the discriminative ability of kinetic and morphologic features. These models were compared to the performance of an expert radiologist. Minimized model was constructed using the most relevant input variables that were determined by the ARD method (automatic relevance determination).

S36

Breast: Imaging, Perfusion and Spectroscopy

Results: Margin type, time-to-peak enhancement and washout ratio showed the highest discriminative ability among diagnostic criteria and comprised the minimized model. The C-indices were the following: 0.863 for the minimized, 0.858 for the maximized, 0.852 for the best kinetic and 0.745 for the morphologic model. Conclusion: A performance of a neural network prediction model is comparable to that of an expert radiologist. Neurostatistical approach is preferred for the analysis of diagnostic criteria when many parameters are involved and complex non-linear relationships exist in the data set.

70 Can combined first-pass perfusion and dynamic contrastenhanced MR mammography improve the differentiation of benign and malignant breast tumors? W. B. Li1, M. H. Li1, G. H. Daniel2, D. H. Wu2, W. T. C. Yuh2; 1Radiology, Shanghai Sixth People's Hospital, Shanghai, CHINA, 2Radiology, University of Oklahoma Health Sciences Center, Oklahoma City, OK. Purpose: To evaluate whether the addition of T2*-weighted firstpass perfusion imaging following dynamic contrast-enhanced MR mammography can increase the differentiation of malignant from benign breast tumors. Subjects and Methods: 65 histopathologically appraised breast tumors, including 37 cases of breast cancer, 13 cases of mastopathy, 11 cases of fibroadenoma and 4 cases of intraductal papillomatosis underwent conventional T1 and T2-weighted, STIR and dynamic contrast-enhenced T1-weighted breast MRI and, subsequently, first-pass, single-section, dynamic T2*-weighted perfusion studies after a bolus injection of 0.1 mmol/kg of body weight Gd-DTPA. For each lesion, morphological features, enhancement patterns, signal intensity parameters were evaluated on a workstation by two radiologists. Interobserver variability was evaluated by the kappa-coefficient. Results: The irregular lesion contour was the most reliable morphological feature indicating malignancy and followd by inhomogeneous contrast enhancement. With T2*-weight perfusion imaging, 28 of 37 carcinomas but only 2 of 28 benign lesions had a signal intensity loss of 18% or more during the first pass. With dynamic contrast-enhanced T1-weighted imaging, 34 carcinomas and 17 benign lesions showed a signal intensity increase of 88% or more in the first image obtained after the administration of contrast material. The signal loss in the T2*-weight perfusion sequence was significantly stronger in the carcinomas than in the benign. Malignant and benign lesions did not differ significantly in any of the quantitatively evaluated signal intensity parameters. Conclusion: The combined use of T2*-weighted first-pass perfusion imaging and T1-weighted dynamic contrast-enhanced MR imaging has the potential to improve the clinical utility of differentiating between benign and malignant breast tumors.

71 Neoadjuvant chemotherapy in breast cancer: the clinical utility of combined DCE-MRI, ADC mapping and proton spectroscopic imaging D. J. Manton1, M. Lowry1, D. Tozer2, A. Maraveyas3, A. Chaturvedi4, J. Greenman5, L. Cawkwell3, M. Lind3, L. W. Turnbull1; 1YCR Centre for Magnetic Resonance Investigations, University of Hull, Hull, UNITED KINGDOM, 2NMR Research Unit, Institute of Neurology, London, UNITED KINGDOM, 3Academic Oncology, University of Hull, Hull, UNITED KINGDOM, 4Clinical Oncology, Hull and East Yorkshire Hospitals NHS Trust, Hull, UNITED KINGDOM, 5Department of Biological Sciences, University of Hull, Hull, UNITED KINGDOM. Purpose/Introduction: Change in tumour volume may be a relatively late manifestation of response to chemotherapy. Studies suggest that using MRI to quantify water apparent diffusion coefficient (ADCW) [1] or microvessel permeability [2], or alternatively using proton MR spectroscopy (MRS) to quantify the water to fat signal ratio (WFR) [3] might provide early indication of ultimate treatment response. Therefore a study was designed to compare these methods in women undergoing neoadjuvant chemotherapy for inoperable primary breast cancer. Subjects/Methods: MRI/S was carried out at 1.5 T (IGE Signa) prior to chemotherapy, between the second and third courses (TP2) and after the final course (TPF) in 25 women following a standard dosage chemotherapy regime. ADCW was measured using EPI (8 gradient weightings, 680 s/mm2 maximum). Microvascularity was assessed using T1-weighted dynamic contrast-enhanced (DCE-) MRI (13 s temporal resolution, 7.5 minute duration) and a 3 compartment pharmacokinetic (PK) model [4]. WFR was measured, at echo times of 30 and 135 ms, using STEAM (water T2 was also estimated). Tumour volume was measured using high resolution 3D, post-contrast, fat-suppressed images and manually traced regions-of-interest. An ultimate decrease in tumour volume (DTVF) of at least 65% was taken to indicate partial response (PR); broadly equivalent to the WHO criterion of a 50% decrease in cross-sectional area. Results: All tumours showed reduced volume at TPF (18 PR; 7 clinically stable, NR). Changes at TP2 in volume, WFR(135ms), T2(water) and PK distribution volume (Ve) all correlated significantly with DTVF. PR and NR data were often sufficiently well separated in order to permit accurate prediction of ultimate response in a substantial number of cases (i.e. with 100% positive and/or 100% negative predictive values, see graph). The proportion of cases where accurate prediction could be made was 40% for volume, 36% for Ve, 24% for WFR(135ms) and 48% for T2(water). The graph also demonstrates how T2(water) and Ve contain complimentary information permitting 100% accuracy when used in combination. Discussion/Conclusion: Results suggest that MRS and DCE-MRI measurements provide complimentary information about early tumour response which can be combined to predict ultimate volume response highly accurately. Prediction of NR could spare patients unnecessary toxicity and permit early transfer to alternative therapy. Prediction of PR could have psychological benefits for patients. 1. Dodd Phys. Med. 13 56 (1997); 2. Button Proc. ISMRM 6 1086 (1999); 3. Jagannathan NMR in Biomed. 11 414 (1998); 4. Tofts Magn. Reson. Med. 17 357 (1991)

Methodology: Human and Animal Models

S37

3. Van Vaals J J et al. ACE: a single shot method for water suppressed localisation and editing of spectra images and spectroscopic images. [1991] Magnetic Res. Med., 19, pp136-190 4. Dendy P P , Heaton B Magnetic Resonance Imaging in ‘Physics for Diagnostic Radiology’. [1991] IOP publishing, Bristol, pp403 5. Matson G B An integrated program for amplitude modulated RF pulse generation and re-mapping with shaped gradients. [1994] Magnetic Resonance Imaging, 12 (8), pp1205-1

Poster Walking Tour Methodology: Human and Animal Models 72 Optimisation and evaluation of the ACE sequence for a phased array breast coil G. M. Egan1, P. Gilligan2; 1Medical Physics, Mater Misericordiae Hospital, Institute of Radiological Sciences, Dublin, IRELAND, 2Medical Physics, Mater Private Hospital, Institute of Radiological Sciences, Dublin, IRELAND. Purpose/Introduction: Using non-invasive MRS techniques, increased amounts of choline are observed in vitro in breast tumours [1]. Pulse sequences currently used are limited for choline detection, as fat and water dominate the choline region of the spectrum [2]. Frequency selective sequences, such as Acquiring Combined Echoes [3], allow suppression of unwanted water and fat signals and thus improve detection of choline in vivo. The purpose of this project was to develop and optimise the ACE sequence for a phased array breast coil and to compare different sequences and coils for performing breast MRS. Subjects and Methods: Phased array coils give improved SNR over single element coils [4]. The ACE sequence has not previously been developed for phased array breast coils. Pulse sequences have been constructed using the Siemens ‘Pargen’ pulse programming language. RF pulses were imported from the ‘Matpulse’ software [5] platform. The optimised ACE sequence has been tested for various parameters including linearity and choline detectability. Results: Results were obtained on the single and phased array breast coils using various sequences, including the optimised ACE sequence. These have been analysed and an optimum protocol for breast MRS developed, outlining recommended coil, sequence and parameters. Discussion/Conclusion: It is possible to implement ACE on a clinical scanner. This gives good water and fat suppression in idealised conditions. However its utility in a range of clinical coils and inhomogeneous tissue in vivo, needs further investigation. References: 1. Barry P A, Mac Kinnon W B, Maylacha P., Gillett D., Russell P. et al. Benign Breast Lesions Distinguished from Invasive cancer by 1H Magnetic resonance spectroscopy on Fine Needle biopsy. [1996] SMR, pp266 2. Ronen S M, Leach M O Imaging biochemistry: applications to breast cancer. [2001] Breast Cancer Res., 3 (1), pp36-40

5:30 pm - 6:30 pm

73 Monocyte infiltration into the brain during experimental allergic encephalomyelitis, as monitored by magnetic resonance imaging E. Blezer1, S. Floris2, E. Döpp2, S. van der Pol2, C. Dijkstra2, K. Nicolay3, H. de Vries2; 1Department of Experimental in vivo NMR, Image Sciences Institute, University Medical Center Utrecht, Utrecht, NETHERLANDS, 2Department of Molecular Cell Biology, VU Medical Center, Amsterdam, NETHERLANDS, 3Department of Biomedical Engineering, Section Biomedical NMR, Eindhoven University of Technology, Eindhoven, NETHERLANDS. Introduction: Increased cerebrovascular leakage and cellular infiltration mark the onset of lesion formation in multiple sclerosis (MS). Recently, ultra small particles of iron oxide (USPIO) have been used to detect the cerebral infiltration of mononuclear cells [1]. Here we combine GdDTPA- and USPIO-enhanced imaging to elucidate the temporal events of Blood Brain Barrier (BBB) opening and mononuclear cell infiltration in an acute model of experimental allergic encephalomyelitis (EAE) in the rat. Methods: EAE (n=40) was induced in 8-11 week old male Lewis rats by immunization with MBP (controls without, n=10). Neurological aberrations were daily graded between 0 and 4 (death) [2]. MRI experiments (4.7 T Varian) were done at day 9, 11, 14 (also controls) and 17, whereafter brains were collected for immunohistochemical characterization. From animals subjected to BBB studies (4 per group), GdDTPA-enhanced T1-weighted maps (% SI-increase due to GdDTPA-leakage) were calculated from two T1W images (TR=650, TE=12.5 ms) before and after a bolus of 0.5mmol/kg GdDTPA (i.v., 10 min in circulation). From animals subjected to USPIO infiltration studies (6 per group), quantitative T2 maps (multi-echo, TR=5000, TE=17.5-87.5 ms) were collected after a bolus of 600 µmol/kg USPIO 7228 (i.v., 24 hrs in circulation, Guerbet, France). Representative ROI's were defined in images in which ventricular pixels were excluded. Percentage areas with abnormal T2-values (T2
S38

Methodology: Human and Animal Models

Results: From 11 days after immunization, USPIO's accumulated in the brain (figure 1), which paralleled with clinical score and monocyte infiltration (assessed by immunohistochemistry). Leakage of GdDTPA for spinal cord and brainstem peaked before clinical score (day 14, ns figure 2). As expected, the cortex showed no abnormalities (figure 2e). Control also showed no aberrations. We used USPIO-enhanced imaging as a tool to evaluate therapeutic approaches and showed that EAE animals (n=10) treated (from day 0) with the immunomodulator lovastatin showed amelioration in clinical score and USPIO-load (data not shown). Discussion/Conclusion: Our results suggest that cerebrovascular leakage and monocytic trafficking into the brain are two distinct processes in the development of inflammatory lesions. Furtermore we showed that USPIO-enhanced imaging is as a valuable tool to monitor therapeutic stategies directed towards monocyte infiltration. References: 1. Douss et al, MAGMA 1999, 8: 185-189; 2. Floris et al, J Neuroimmunol 2002, 127: 69-79.

74 NMR microscopy of hyper-stressed plant roots I. Kaufmann1, L. Wegner2, U. Zimmermann2, A. Haase1; 1Physikalisches Institut, EP5, Universitaet Wuerzburg, Wuerzburg, GERMANY, 2Institut fuer Biotechnologie, Universitaet Wuerzburg, Wuerzburg, GERMANY. Introduction: NMR microscopy is an ideal method for studying the internal structural changes of plant roots under stress conditions and allows one to identify irreversible damages. Subjects and Methods: The hardware used was a vertical Bruker 11.75T AMX500 spectrometer, a custom-built probehead and a Helmholtz coil (R=5mm) fitted around a dedicated NMR-glasstube for plant roots.The root of a Zea mays plant was allowed to grow into the glass tube, containing 4.0mmol NO3 nutrient solution, for 8 days after germination. To minimize T2* effects (due to root aeration) and to achieve a high time resolution for functional studies, we used a Turbo Spinecho Method (TSE) combined with saturation recovery for T1 maps (TE=8.7ms, 256*256 points, FOV= 6,67mm * 6,67mm, TSE with 8 echo train, duration = 28min). At time index 17.67h the nutrient solution was removed and refilled at 34.25h. The root was observed with 72 consecutive T1 maps over about 40h. Results: Fig. 1 and 2 show portions of T1 and spin density parameter maps of the root and a reference capillary (in 1a the nutrient solution was already partially removed and later on filled to right below the examination area). Although a swelling of the root could be observed, the plant did not survive, as indicated by wilting. Just above the examination area this root was completely dehydrated. Plot 3 and 4 show the shrinkage of the root as well as T1 and spin density variation over time for certain root areas. Discussion: The central cylinder (r=0.0mm to r=0.2mm in fig. 3 and 4) and the cortex (0.2mm to 0.5 mm) of the root shrank evenly whereas the re-swelling of the cortex took place with a delay of 5 hours. Additionally, T1 remained close to 0.7s within the cortex compared to values above 2.0s prior to the dehydration. This may indicate the destruction of the cortical cells. A dark granulated ring in the outer part of the central cylinder (see fig. 2 d and e) indicates that parts of the protoxylem or phloem could not be refilled after an irreversible dehydration. We were able to show that hyperstressed root tissue differs from fully functional tissue in its NMR parameters, which would allow us to determine a dehydration limit in further studies.

Methodology: Human and Animal Models 75 High Field Magnetic Resonance Microscopy of pig cartilage J. C. Tsai, T. Weber, T. Neuburger, A. Purea, V. Behr, A. Hasse, A. Webb; Physik, Wuerzburg University, Am Hubland, GERMANY. Introduction: 1H MRI is a promising method in detecting early degenerative changes in cartilage before volume loss and clinical symptoms develop in osteoarthritis. This would be particular valuable in early diagnosis and therapy evaluation for osteoarthritic patients. Prior works have pointed to several directions for possible early MR indicators of degenerative changes, e.g., the detection of striations in the radial zone, Na+ concentration measurements, T1 maping with Gd(DTPA)2-, T2 mapping, and diffusivity imaging. Nevertheless, progress for validation has been hampered by a shortage of high resolution MR microscopy images to correlate with histologic findings (1). We present our first attempt with MR microscopy at 17.6 T with rapid high resolution imaging of healthy pig femoral condylar cartilage. Objects and Methods: MR imaging: Imaging experiments were performed on a Bruker Avance 750 WB system at a 750 MHz proton resonance frequency. The probe contained a 38 mm linear birdcage resonator. Optimal contrast was achieved using a 2D multi slice turbo spin echo sequence with 16 echoes and an effective echo time (TE) of 19.4 ms and a repetition time of 3s. In a total scan time of less than 7 minutes a spatial resolution of 117x117 µm² in-plane and a slice thickness of 350 µm was achieved. Pig cartilage: A knee specimen from a healthy pig was obtained. The knee specimen was dissected and medial femoral condylar cartilage was excised and sectioned to fit into an NMR tube. It was placed in an NMR tube with saline. Results: The four cartilage layers, the low intensity superficial layer, intermediate signal intensity transitional layer, low intensity, heterogeneous radial layer, and intermediate intensity deep layer could be clearly distinguished with high contrast on the T2 weighted MR images. These layers correspond to the four histologic cartilage zones, the surface zone, transitional zone, radial zone, and calcified zone. Some vertically oriented striation pattern is visible in the radial zone. Discussion: Vertical striation pattern in the radial zone occurs due to T2 variations in the cartilage. We were able to demonstrate MR microscopy can be achieved with good resolution at 17.6 T with a scan time of just 7 minutes compared to hours in earlier works. Next steps in our study include optimized T2 weighted imaging to accentuate striation pattern in the radial zone, Na+ imaging, and DTI for histologic correlation.

References: 1. DW Goodwin, JF Dunn, Topics in Magnetic Resonance Imaging, 9(6):337-347. 1998.

S39

76 Water suppression for short-TE MRSI in 1.5 T using short trains of hyperbolic secant RF pulses Z. Starcuk, Jr., J. Starcukova, J. Horky, Z. Starcuk; Department of NMR, Institute of Scientific Instruments, ASCR, Brno, CZECH REPUBLIC. Purpose: Efficient water suppression (WS) insensitive to B0 and B1 inhomogeneities and to the dispersion of water relaxation times T1 and T2 is a basic challenge in short echo time (TE<35 ms) spectroscopic imaging, especially in lower fields (<3 T) and if detection of metabolite signals close to the water resonance (0.5 ppm) is required. Most often, a preparatory period with several frequency-selective amplitude-modulated (AM) RF pulses is employed (1-3). However, the use of AM pulses might be problematic at 1.5 T because of relaxation during the very long RF pulses needed for the desired frequency selectivity. The resulting excitation might exhibit significant reduction of the WS efficiency and corruption of the intensities of metabolite signals close to the water resonance. Sequence length or unsuitable WS profiles belong to the drawbacks of techniques with pulse trains of frequency swept (FS) RF pulses (4-6). Method: WS sequences with only small (~4) numbers of FS hyperbolic secant (HS) RF pulses were developed using numerical solutions of the Bloch equations including relaxation. The novelty of these sequences consists in the way the relaxation effects are compensated by utilizing HS pulses with alternating frequency sweep directions. Results: Several sequences with various pulses and timing have been tested. Fig. 1 shows the calculated excitation profiles of comparable sequences with (a) four 140-ms HS FS pulses, (b) 50-ms P10, (c) 25-ms gaussian AM pulses. T1=1000 ms, T2=80 ms were assumed. Asymmetric profiles with a reduced transition bandwidth can also be constructed. The insensitivity of method (a) to T1 is documented by Fig. 2. Conlusions: WS sequences with a few long HS pulses with alternated sweep directions may provide WS regions with flat plateaus and narrow transition regions, with low B1 and relaxation time sensitivity, which could not be achieved with AM RF pulses only. Such an approach appears suitable for lower field strengths, where relaxation effects severely affect the excitation profiles. References: 1. Haase A, Frahm J, Hänicke W, Matthei D [1985] Phys.Med.Biol. 30:341. 2. Ogg RJ, Kingsley PB, and Taylor JS [1994] J. Magn.Reson. B 104:1-10. 3. Tkác I, Starcuk Z, Choi I-Y, Gruetter R [1999] Magn.Reson.Med. 41:649-656. 4. deGraaf RA, Luo Y, Garwood M, Nicolay K [1996] J.Magn.Reson. B 113:35-45. 5. deGraaf RA, Nicolay K [1998] Magn.Reson.Med. 40:690696. 6. Starcuk Z Jr. et al. [2001] J.Magn.Reson. 152:168-178.

S40

Methodology: Human and Animal Models changes during an antibiotic treatment. We developped a rapid CINE type acquisition to shorten the experiment duration. Subjects and Methods: We developed a mouse lung inflammation model with Klebsiella pneumoniae. Mice are infected intratracheally at doses ranging between 10e5 and 10e8 bacteria per animal. Images of the lungs are acquired 24 hours post-infection and ervery day to day 3. Antibiotics consists of ampicillin/gentamycin IM injections. In order to optimise the MRI parameters for visualisation of lung inflammation, we created a severe inflammation in mice following IV administration of oleic acid. DOTA was used as a contrast agent using dose of 0,3 to 1 mmmol of DOTA per kg in bolus. The MRI experiments were carried out on a 7T imaging spectrometer (Varian, Inova) equipped with an 11cm available bore 120mT/m gradient coil and a birdcage custom made probe of 3.6 internal diameter. 2D multislice short TE CINE gradient echo sequence were used with T1 weighting. (TR/TE : 0.35/2.9ms, slice thickness of 0.5 mm, FOV = 3.5*3.5cm2, 512*256 acquisition points, with a shifted echo in the readout dimension) with an ECG gated system. High field induces problems of reduced T2* especially in the case of inhomogeneous parenchyma tissue, however they were limited by using short TE. The total duration time was about 10min. Histology and image processing were achieved. Results: We observed a high intensity signal coming from the edema regions. The injection of the contrast agent enhances it by 40%. A local increase of about 16% to 48% was observed in the parenchyma in the KP model, as confirmed by the histology. Effects of the antibiotics were also quantified. The MRI results were coherent with the histopathological data. Discussion/Conclusion: We have demonstrated the feasibility of studying inflammation on mice lungs in vivo by MRI at 7T on an infectious disease by Klebsiella pneumoniae bacteria and after treatment with an antibiotic.

77 In vivo mice lungs MRI at 7T for the study of infectious diseases evolution B. Doan1, R. Tournebize2, P. Sansonnetti2, J. Beloeil3; 1Cnrs, Laboratoire de RMN Biologique, Gif sur Yvette, FRANCE, 2Institut Pasteur, Laboratoire de Pathogénie Microbienne, Paris, FRANCE, 3Cnrs, Centre de Biophysique Moléculaire, Orléans, FRANCE. Purpose/Introduction: MRI of normal lung parenchyma suffered from low signal intensity as motions artefacts from breathing and cardiac pulsation and the high magnetic susceptibility difference within lungs destroy the MRI signal. However, adapted sequence based on short echo time gradient echo improved the quality of the images. However, we can benefit from this inconvenient with a pathology generating an edema, the appearance of the signal could be then detectable. Our aim is to visualize temporally and spatially the development of inflammation consecutive of a bacterial infection by MRI and its

78 Slice-selective coherence transfer for in vivo simultaneous 3D spatial localization and doubly selective spectral editing of GABA in the human brain J. Shen; Mib, NIMH, Bethesda, MD. Introduction: Symmetric, linear phase, selective pulses (e.g., sinc pulses) is configured to perform well-defined slice-selective coherence transfer such as polarization transfer and conversion of multiple quantum states. The application of these pulses leads to simultaneous 3D localization and doubly selective spectral editing of neurotransmitter GABA in each single shot. Methods and Results: Fig. 1 shows simulated spatial profiles of the polarization transfer yield and the double quantum yield using the proposd 90o slice-selective universal rotator RF-gradient pulse scheme. For 2IxSz → 2IzSx and 2IxSz + 2IzSx → DQCy, ~100%

Methodology: Human and Animal Models polarization transfer yield and double quantum yield are obtained respectively within the passband. The GABA editing pulse sequence is depicted in Fig. 2. It uses a doubly selective DANTE pulse for GABA-3, 4 DQ preparation. The second 90o pulse is configured to act as a slice-selective universal rotator pulse for conversion of antiphase single quantum coherence into double quantum coherence along the z direction. After DQ labeling, the GABA-4 is rephased by a pair of identical 3.5 ms hyperbolic secant pulses (µ = 5, 1% truncation) along the y direction. Fig. 3 shows the in vivo spectra with TE = 68 ms from a 4 x 2 x 3 cm3 voxel in the occipital lobe of a healthy volunteer obtained at 2.1 Tesla. Top trace: the unedited spin-echo spectrum; NS = 8. Bottom trace: edited spectrum; NS=128. The doubly selective double quantum filter suppressed the dominant creatine singlet at 3.0 ppm and revealed the underneath GABA-4 doublet at 3.0 ppm. The doubly selective double quantum filter also suppreses other singlets and coupled spins such as choline, NAA, glutathione, glutamate and macromolecules. Discussion: In our previous GABA editing method using a doubly selective DANTE refocusing pulse for selective preparation of GABA-3, 4 double quantum spatial localization was achieved using the 8-step three-dimensional ISIS method. In that sequence only the first 90o pulse and the second 180o pulse are available for conventional spatial localization because the doubly selective DANTE pulse is necessary to suppress overlapping glutathione and mobile macromolecules around 3.0 ppm. Application of the newly proposed slice-selective universal rotator pulse has made it feasible to achieve simultaneous doubly selective spectral editing and 3D spatial localization of GABA in a single shot.

S41

79 In vivo follow-up of mouse spinal cord injury by diffusion MR microscopy J. Bonny1, M. Gaviria2, J. Donnat1, H. Haton2, A. Privat2, J. Renou1; 1Stim, INRA, Saint-Genes Champanelle, FRANCE, 2U583, INSERM, Montpellier, FRANCE. Purpose: Therapeutic intervention soon after spinal cord injury (SCI) is crucial for controlling the secondary injury extent and improving functional recovery. If most of the main actors of neural toxicity after trauma have been identified, clinical efficacy of therapeutic tools developed in the clinical context remains often marginal. The absence of significant results for clinical trials raises the question of pre-clinical predictiveness. Temporal dimension of pathophysiological changes after SCI being still not well known, therapeutic windows are roughly estimated and it is difficult to proceed to a pharmacological intervention providing optimal results (pre-clinical and clinical domains). Our objective was to follow up the pathological process in the spinal cord of living mice using MR microscopy as soon as one hour after an experimental SCI. Subjects and Methods: Experiments were performed on an Avance DRX400 micro-imaging system. Mice were anaesthetised in an animal handling system with the thorax centred in a 20-mm diameter birdcage coil. Due to their sensitivity for monitoring injuries, diffusion-weighted images (DWI) were collected perpendicularly to the spinal cord at the 8th thoracic level (T8) with a field of view of 12x12 mm2, an image matrix of 96x96 and a 1-mm slice thickness. The thoracic location of the lesion was mandatory, since it yields a well characterized deficit of the hindlimbs and does not hamper vital prognosis. DWI were produced by a PGSE sequence with diffusion sensitising gradients applied in parallel and perpendicularly to the WM fascicle. To minimize bulk motion effect, the sequence was triggered from every breath-in using a dedicated respiratory sensor. Using 16 averages, a typical mouse breathing frequency of 100-bpm led to an acquisition time of ~15-min. Results: We show the first DWI depicting the thoracic spinal cord of living mice with a high contrast between WM and GM. Following SCI, lesions are clearly discernable with an early thickening of the dorsal horns followed by a progressive loss of GM/WM contrast. Discussion/Conclusion: Despite both the narrowing of the spinal cord diameter at T8 and the extreme sensitivity of DWI to physiological motion, the obtained temporal and spatial resolutions are sufficient to highlight the physicochemical changes in the injured tissues.

S42

Methodology: Human and Animal Models

Our MR protocol provides a tool to improve our knowledge of the anatomical and functional time-related extent of the injury. Additionally, it represents a powerful tool for the phenotyping of transgenic mice models used in spinal cord research.

Knopp MV, Larsson HBW, Lee T-Y, Mayr NA, Parker GJM, Port RE, Taylor J, Weisskoff RM [1999] J.Magn.Reson.Imaging 10:223-232

80 Technique of pixelwise comparison of microvessel-density and model parameters of dynamic MRI M. Heilmann1, F. Kiessling2, S. Vosseler3, N. E. Fusenig3, L. R. Schad1; 1Biophysics and Medical Radiation Physics, Deutsches Krebsforschungszentrum, Heidelberg, GERMANY, 2Oncological Diagnostics and Therapy, Deutsches Krebsforschungszentrum, Heidelberg, GERMANY, 3Differentiation and Carcinogenesis in vitro, Deutsches Krebsforschungszentrum, Heidelberg, GERMANY. Introduction: Microvessel-density (MVD) is an important factor in characterizing tumor vascularisation. The presented technique is a tool to compare non-invasively acquired parameter maps of dynamic contrast-enhanced MRI with MVD obtained by immuno-fluorescence-analysis. By this a verification whether dynamic MRI-parameters reflect MVD can be performed. Subjects and Methods: In a pilot experiment three nude mice with untreated tumors of the cell-line HaCaT-ras-A-5RT3 were examined by dynamic MRI in a clinical 1.5T MR-system (Vision, Siemens) using a custom-made animal-coil and a T1w-GE-sequence. One coronar tumor slice was measured with 60 repetitions at a spatial resolution of (0.5x0.5x2.0)mm3 and a temporal resolution of 8s. 0.4mmol/kg Gd-DTPA were injected as a bolus into the tail vein [1]. For HE-histology and immuno-fluorescenceanalysis the tumor was cryo-fixed and dissected in the same orientation as the MR slice. For MVD-determination anti-CD31staining was applied. Data analysis of the dynamic MRI data was done by pixel-wise fitting of the signal-time-course to a two-compartment-model [2] in order to determine amplitude A and rate constant kep. MVD-measurement was performed in three sections per tumor at a spatial resolution of (0.4x0.3x0.006) mm³. Color-coded maps were calculated for both the modelparameters A and kep and the MVD. Color-coding ranged from 0 to MAX in 256 steps (fig.1). Results: In the color-coded maps of the dynamic MRI parameters typical values of MAX were found holding for all three tumors: MAX(A)=1.5 and MAX(kep)=5.0 (fig.2). But in the color-coded maps of MVD the three sections of the same tumor had to be scaled differently: MAX(#1,#3)=12.0 and MAX(#2)=6.0 (fig.3). Additionally the cryo-fixed sections were deformed because of central necrotic regions. Conclusion: By using cryo-fixation only a qualitative comparison in terms of visual comparison of the parameter-maps is possible. To perform a quantitative pixelwise correlation between MRI-parameter- and MVD-maps a fixation of histological specimens in paraffin wax is recommended. Further MVD has to be normalized to overcome fluctuations of brightness in-between the histological sections leading to different MAX-values in the MVD-maps. Then 3-5 immuno-fluorescence-sections can be added up to cover the same slice thickness as in dynamic MRI. References: 1. Kiessling F, Heilmann M, Vosseler S, Lichy M, Krix M, Fink C, Kiessling I, Steinbauer H, Schad L, Fusenig NE, Delorme S [2003] Int.J.Cancer 104:113-120 2. Tofts PS, Brix G, Buckley DL, Evelhoch JL, Henderson E,

81 Characterization of myocardial perfusion defects in GotoKakizaki rats, a type II diabetes model using a spin-labeling MRI method I. Iltis1, F. Kober1, M. Desrois1, C. Lan1, B. Portha2, P. J. Cozzone1, M. Bernard1; 1Centre de Résonance Magnétique Biologique et Médicale UMR CNRS 6612, Faculté de Médecine, Marseille, FRANCE, 2Laboratoire de Physiopathologie de la Nutrition UMR CNRS 7059, Université Paris 7 / D. Diderot, Paris, FRANCE. Type II diabetes is a non-insulin dependent diabetes mellitus (NIDDM). Diabetic patients are exposed to a higher frequency of cardiovascular events, in particular because of microvascular alterations due to endothelial dysfunction. The Goto-Kakizaki (GK) rat is a non-obese model of NIDDM analogous to the early stages of diabetes in humans (1). This model is then particularly suited to study cardiovascular disorders, such as myocardial perfusion defects (2). An ECG- and respiration-gated gradient-echo spin-labeling MRI method was used to evaluate myocardial perfusion in vivo in Wistar and GK rats. This method is non-invasive (no contrast agent injection), quantitative and adapted to high heart rates, allowing the use of isoflurane anesthesia. Short-axis perfusion maps were obtained at rest with a high in-plane resolution (234x468µm2). Adult female Wistar control (8 months, n = 8) and GK (8 - 13 months, n = 11) rats were anesthetised with isoflurane (2 to 2.5 %) under spontaneous respiration. Total experiment time was about 25 minutes. The figure shows typical perfusion maps obtained from control (left) and GK (right) rats. Scale is given in ml/g/min. GK female rats show a significant decrease in perfusion compared with controls (4.0 ± 1.3 vs. 5.8 ± 1.6 ml/g/min, mean ± SD, p=0.016). We report for the first time that perfusion is altered in vivo in a nonobese NIDDM model. The GK rats used here are a chronic model of untreated type II diabetes. This method provides a useful tool to assess the potential improvement of myocardial perfusion by antiischemic or anti-diabetic drugs. 1. F. Picarel-Blanchot, C. Berthelier, D. Bailbe and B. Portha [1996] Am. J. Physiol. 271: E755-62. 2. Z. J. Cheng, T. Vaskonen, I. Tikkanen, K. Nurminen, H. Ruskoaho, H. Vapaatalo, D. Muller, J. K. Park, F. C. Luft and E. M. Mervaala [2001] Hypertension. 37: 433-9. This work was supported by the program “Imagerie du Petit Animal” (2001-2002).

Abdomen and Pelvis: Imaging

S43

82 Diffusivity-microimaging of arteriosclerosis at 3 Tesla A. Berg1, J. Sailer2, T. Rand2, E. Moser1,3; 1Department for Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Department of Diagnostic Radiology, Section Angiography and Interventional Radiology, Vienna, AUSTRIA, 3Department of Diagnostic Radiology, University of Vienna, Vienna, AUSTRIA. Introduction/Purpose: A major challenge of MRI of atherosclerotic plaques is represented by the spatial resolution, which is needed to detect the small pathologic alterations below about 500x500x2000µm3 obtained in clinical routine diagnosis. Due to the higher signal-to-noise-ratio (SNR) microimaging has mainly been performed on high-field (B > 4,7 T) small-bore MR-microscopy systems. We present methods and results for obtaining diffusivity micro-images on a 3T MR-scanner at 109x109x1000µm3 pixel-size resolution. We investigate the possibility for the discrimination of atherosclerotic plaques by diffusivity as contrast parameter and referencing to histological images. Advantages and limitations of diffusivity microimaging at 3T are discussed. Materials/Methods: High-resolution images were obtained on a 3T MR-scanner (BRUKER MEDSPEC S300, free inner diameter: 55 cm). For strong diffusion-weighing we used an additional, actively-shielded gradient system (Gmax=200mT/m). A small detector (15 mm i.d. cylindrical resonator) was required to ensure adequate SNR. For diffusion-weighting, a pulsed gradient field sequence was used, along with stimulated echoes. Diffusivity (Dapp) maps were obtained from an exponential fit for each pixel element of the four differently diffusion-weighted images (bmax=2007 s/mm2) according to the equation: S/S0=exp[Dappγ2δ2(Delta-1/3δ)g2] (diffusion time: Delta=23.05ms, gradientduration: δ=6ms). Short sections of blood vessels were excised from 6 males and 7 females (mean age:68y) diagnosed with acute stenosis. Specimens were taken from the iliac, carotid and femoral arteries. After MRscanning 2-3µm thick, haematoxylin stained slices were cut from paraffin blocks and interpreted by a pathologist concerning types of plaques and tissue (histology). Results: Out of several different slices and specimens we select an example exhibiting a rather complex morphological plaque structure. Despite the much thicker slice used for the diffusivity map (fig. 1), most of the structures seen in the histological image can be identified. The observed apparent diffusivities range from about 2x10-4 mm2/s (dark) for coagulated blood to D=1.4x10-3 mm2/s (hyperintense) in collagen rich tissue.

Figure 1. Diffusivity map of an atherosclerotic artery (Vs:109x109x1000µm3). Adventitia, media, lumen and pathologic degenerations: reorganized thrombus, coagulated blood in the lumen, collagen and lipid rich tissue, can be observed. Conclusion: High-resolution parameter-weighted MRI and parameter-imaging at about 110µm pixel resolution has been obtained for excised tissue on a 3T MR-scanner using a sensitive resonator and a powerful gradient system. Diffusion mapping is possible at sufficiently high spatial resolution to discriminate different atherosclerotic plaques. No high-field small-bore MR-system is necessary for MR-microimaging on excised tissue though still advantages due to the higher SNR are present.

Poster Walking Tour Abdomen and Pelvis: Imaging 5:30 am - 6:30 am

83 Detection and characterization of hepatocellular carcinomas with Gd-BOPTA-enhanced magnetic resonance imaging (MRI) and Multislice-spiral computed tomography (CT): are they interchangable? I. Sansoni, A. Laghi, I. Baeli, R. Ferrari, D. Marin, M. Celestre, R. Passariello; Radiological Sciences, Policlinico Umberto I, University of Rome, Rome, ITALY. Purpose: We prospectively compared the diagnostic accuracy of Gd-BOPTA-enhanced magnetic resonance imaging (MRI) and tetraphasic mutlislice-spiral computed tomography (CT) in the detection and characterization of hepatocellular carcinoma. Subjects and Methods: Sixty patients with chronic hepatitis underwent liver CT and MRI examination, within a mean interval of one month. We included patients with either a new non-cystic liver lesion at sonography or increased α-FP levels (>40µg/l). MR examinations were performed on a 1.5T scanner. The acquisition protocol is the following: pre-contrast T2-weighted HASTE (TR=inf TEeff=90 FA=180° ETL=104 Thickness=6mm Gap=20% Matrix=192x256 SA=1 N.slices=20 TA=20) and T1-weighted

S44

Abdomen and Pelvis: Imaging

FLASH (TR=156msec TE=4.35msec FA=80° Thickness=6mm Gap=20% SA=1 N.slices=20 TA=16); dynamic contrast-enhanced T1-weighted FLASH acquired at the arterial, portal and equilibrium phases; delayed (60 min) T1-weighted FLASH hepato-biliary phase. CT studies were performed using a 4-slice multidetector scanner with a tetraphasic acquisition protocol (unenhanced, arterial, portal venous and equilibrium phases) following dynamic injection of non-ionic contrast medium at a flow rate of 4.0 ml/sec. Two radiographers performed a blind evaluation of MRI and CT considering the presence, the number and the characterisation of lesions. Gold-standard of our image results were histopathological studies of the surgical specimens and/or follow up imaging. Results: No statistically significant difference in lesion detection (p>0.05) was observed between MR and CT studies, as GdBOPTA-enhanced magnetic resonance imaging detected 51/56 lesions (27 ≥ 2 cm and 24 < 2 cm) and Multislice spiral CT detected 52/56 hypervascular focal liver lesions (25 ≥ 2 cm and 27 < 2 cm). MRI showed correct lesion characterization in 52 cases and multislice CT in 39 cases. A statistically significant difference between CT and MR was observed for characterization of focal lesions smaller than 2 cm. Conclusion: In terms of lesion detection rate, no statistically-significant difference between Gd-BOPTA-enhanced magnetic resonance imaging and multislice-spiral computed tomography was observed, independently of lesion size. If lesions smaller than 2 cm are considered, Gd-BOPTA MRI provided additional clues for lesion characterization.

84 Gd-BOPTA versus SH U 555 A – two different MR contrast agents for the detection and differential diagnosis of hypervascular liver tumors G. Schneider, R. Seidel, P. Fries, W. Loytved, B. Kramann, F. Ahlhelm; Dept. of Radiology, Universitaetskliniken des Saarlandes, Homburg/Saar, GERMANY. Purpose: Gadobenate dimglumine (Gd-BOPTA MultiHance, Bracco Imaging SpA, Milan Italy) and SH U 555A (Resovist, Schering AG, Berlin, Germany) are MR contrast agents recommended for both dynamic and delayed phase imaging of the liver. The study was conducted to compare these two agents for the detection and characterization of hypervascular hepatic lesions. Material and Methods: 32 patients with hypervascular hepatic tumors (15 HCC, 8 FNH, 1 NRH, 8 metastases) each received GdBOPTA (0.05 mmol/kg BW) and SH U 555A (8 µmol Fe/kg) with an interval of 3-5 days between examinations. Unenhanced T1w and T2w images and enhanced T1w images in the arterial and portal-venous phase were acquired for both contrast agents. Additionally, T1w images were acquired at ~90 min after GdBOPTA administration while T2w and T1w images were acquired at ~15 min after SH U 555A administration. Two independent blinded readers evaluated the number, size and DDX of lesions. Results: Readers 1 and 2 detected 61 and 65 lesions, respectively, in the 32 patients. In two patients with cirrhosis three well-differentiated HCC were not detected after SH U 555A but were delineated clearly on dynamic imaging after Gd-BOPTA. Small (< 1cm) FNH were not seen on SH U 555 A enhanced images in two patients with multiple FNH. Lesion characterization was preferred on dynamic images after Gd-BOPTA due to greater increases of signal intensity (SI). Only slight increases of lesion SI were noted in the arterial phase after SH U 555 A at the recommended dose and this did not permit an accurate diagnosis in the majority of lesions. Distinction

between cystic and solid, perfused lesions was possible, however, after SH U 555 A. Conclusion: Delineation of liver tumors is possible with both contrast agents although greater confidence for detection and characterization is achieved with Gd-BOPTA.

85 MR imaging of hepatic apoptosis with Gd-EOB-DTPA C. Laumonier, S. Boutry, J. Segers, L. Vander Elst, R. N. Muller; Dpt Organic Chemistry/NMR Lab, University of Mons-Hainaut, Mons, BELGIUM. Purpose/Introduction: Hepatospecific MRI contrast agents such as Gd-EOB-DTPA Eovist® Schering, Germany) are known for their capacity to differentiate healthy from tumorous hepatocytes (1). Eovist® is therefore commonly used to detect hepatic tumors. Other pathologies can also be detected using Gd-EOB-DTPA, for example, biliary obstruction and chemically induced hepatitis. This study aims at the detection of apoptosis in the liver with GdEOB-DTPA. Apoptosis, a type of cellular death, intervenes in various pathologies and happens consecutively to anti-tumorous treatments. Subjects and Methods: Apoptosis is induced in the liver by intravenous injection of anti-fas antibody (2) on balb C mice. Induction of apoptosis was confirmed by the TUNEL histological technique (TdT mediated dUTP nick end labeling). In vivo and ex vivo experiments were performed on healthy and apoptotic mice. In the case of ex vivo experiments, the livers were perfused at a flow rate of 2 ml/min (closed circuit (150 ml) during 30 min and open circuit during the 30 min rinsing period) with a Krebs-Henseleit buffer containing 0.4 mM of Gd-EOB-DTPA. In vivo, the contrast agent was administered by intravenous injection (0.03 mmol/Kg) in the caudal vein. MRI was performed on a Bruker Avance-200 (4.7 T) system. Quantitative analysis of MRI data, texture analysis and color mapping of MR images were respectively carried out with Paravision, MAZDA and OSIRIS programs. Gd concentration were measured by ICP-AES (Inductively coupled plasma atom emission spectrometry) on mineralized livers removed 1 h after injection of Gd-EOB-DTPA. Results: 2 h after intravenous injection of anti-fas, histology shows that more than 80% of hepatocytes are apoptotic and distributed as islets. On in vivo MR images collected 1 h after injection of GdEOB-DTPA, apoptotic livers appear more granulous than healthy ones. Texture analysis quantitatively confirms this feature. Ex vivo, signal intensity and ICP-AES measurements suggest that apoptotic livers are taking up less contrast agent than healthy livers. Discussion/Conclusion: Hepatospecific contrast agents can be useful for the detection of apoptotic processes. The distribution of the affected areas is depicted by the image intensities which reflect the inability of apoptotic cells to internalize the paramagnetic molecule. Reference: 1. Lewin M, Clement O, Belguise-Valladier P, Tran L, Cuenod CA, Siauve N, Frija G [2001] Invest Radiol. 36(1):9-14 2. Ogasawara J, Watanabe-Fukunaga R, Adachi M, Matsuzawa A, Kasugai T, Kitamura Y, Itoh N, Suda T, Nagata S [1993]Nature 364:806-809

Abdomen and Pelvis: Imaging 86 The optimal timing of delayed Gd-BOPTA Enhanced MRI: normal versus cirrhotic livers I. Sansoni, A. Laghi, P. Paolantonio, F. Iafrate, M. Celestre, D. Marin, R. Passariello; Radiological Sciences, Policlinico Umberto I, University of Rome, Rome, ITALY. Purpose: The aim of our study was to define the correct delayed timing (40, 120, 180 minutes) of Gd-BOPTA enhanced MR Imaging of the liver in patients with normal hepatic function and in patients with cirrhosis. Subjects and Methods: Seventeen patients (7 cirrhotic and 10 normal) with different hepatic lesions were included in our study, after their informed consent. MR study was performed on a 1.5T scanner. The acquisition protocol was the following: pre-contrast T2weighted HASTE (TR=inf TEeff=90 FA=180° ETL=104 Thickness=6mm Gap=20% Matrix=192x256 SA=1 N.slices=20 TA=20) and T1-weighted fat-sat VIBE (TR=5,2 TE=2,6 FA=20° SlabThick=150 Eff.Thick=5mm n.partitions=30 Matrix=104x256 TA=18). After i.v. administration of Gd-BOPTA (0,2 mM/kg) dynamic T1-weighted fat-sat VIBE at arterial (15 seconds), portal venous (70 seconds) and equilibrium phases (150 seconds) were obtained; delayed T1-weighted fat-sat VIBE hepato-biliary phases were acquired at 40, 120, 180 minutes after i.v. administration of Gd-BOPTA. Signal intensity (SI) measurements of the hepatic parenchyma, bile and lesions were recorded for each of the GdBOPTA-enhanced acquisitions for all patients. Several parameters were calculated from the measured SI, such as contrast-to-noise ratio (CNR) of liver, bile and lesions using basal liver SI as the reference. We also defined the percentage of liver, bile and lesion SI changes in the different delayed phases compared to basal condition. Finally we compared CNR and percentage of cirrhotic patients with those of patients with normal liver function. Results: Cirrhotic patients were found to have a later concentration of bile (the peak of concentration was at 180 minutes), probably due to reduced hepatic function; moreover, non-cirrhotic patients had higher signal intensity of bile (an average of 105 vs. 45) at the peak phase. Patients with cirrhosis also showed to have lower and late enhancement of hepatic parenchyma. Conclusion: Considering CNR and SI values of liver, bile and lesions, patients with cirrhosis should undergo delayed sequence acquisition 180 minutes after Gd-BOPTA i.v. injection, whereas for non-cirrhotic patients 120 minutes is the optimal delayed interval.

87 Magnetic Resonance Imaging to investigate the influence of posture on gastric physiology R. Treier1, A. Steingoetter1, D. Weishaupt2, P. Boesiger1, M. Fried3, W. Schwizer3; 1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, SWITZERLAND, 2Department of Radiology, University Hospital Zurich, Zurich, SWITZERLAND, 3Department of Gastroenterology, University Hospital Zurich, Zurich, SWITZERLAND. Purpose/Introduction: The influence of posture on the physiology of the human stomach has so far been analyzed exclusively using γscintigraphy. In this study a method was developed and evaluated to assess gastric emptying, relaxation, motility and gastric meal distribution in sitting position (SP) and right decubitus position (RDP) using two different MR systems. Subjects and Methods: Gastric function of four volunteers was monitored in SP using a 0.5 T open MR system and in RDP using a

S45

1.5 T compact MR system on two different days. The subjects ingested the solid phase of a low fat mixed solid/liquid meal followed by the liquid phase either in SP or RDP. Repeated sequences of a volume scan followed by a motility scan were performed in regular intervals. Total, proximal, distal meal and stomach volume and contraction frequency and velocity of gastric peristalsis were calculated from the MRI data. Results: No difference in gastric emptying pattern was observed between SP and RDP. However, faster emptying of proximal meal volume was observed in RDP after t = 30 min (AUC of proximal meal volume from t = 30 min to t = 80 min: SP, 35 ± 3; RDP, 31 ± 3; p < 0.05). This resulted in a divergence of the overall emptying curve following this lag time. There was no difference in proximal and distal gastric relaxation between SP and RDP after total meal ingestion. However, in both positions relaxation was different for proximal and distal stomach after solid (relative change in stomach volume: SP, prox. (40 ± 4)% vs. dist. (4 ± 2)%, p < 0.01; RDP, prox. (30 ± 10)% vs. dist. (6 ± 5)%, p < 0.05), but not after liquid meal ingestion. Gastric air volume was lower in SP than in RDP during the entire study period (SP, (10 ± 6)%; RDP, (37 ± 16)%; p < 0.05). Contraction frequency and velocity of gastric peristalsis showed no posture dependency. Conclusion/Discussion: In this study, MRI provides new insights into gastric physiology by the simultaneous assessment of the determining parameters of gastric function and the analysis of the posture dependency of these parameters. The data suggest that posture has effects on gastric physiology regarding gastric emptying patterns and gastric air volume. In contrast, gastric motility was not affected by posture. These posture dependencies must be considered in the interpretation of data measured in lying position using compact clinical MR systems.

88 MRI of the small bowel using a Polyethilenglycol solution as an oral contrast agent in a population with celiac disease P. Paolantonio, A. Laghi, M. Celestre, D. Marin, F. Iafrate, I. Baeli, R. Passariello; Department of Radiology, University of Rome, Rome, ITALY. Pourpose: The aim of our study was to describe morphological abnormalities of the small bowel in a population of patients with known celiac disease using MR Imaging with polyethilenglycol solution as an oral contrast agent (PEG-MRI). Materials and Methods: Twentynine patients, 19 adults and 10 children with known celiac disease underwent MR study of the small bowel. After an overnight fast, immediately before MR examination a fixed amount of 10 ml/kg of body weight of PEG solution was orally administered. MR study protocol included HASTE (TR/TE/acq.t.: inf/90ms/18s) and True-FISP (TR= 4,8 ms; TE= 2,3 ms; Flip Angle 50°). sequence obtained sequentially on axial and coronal planes. Results: Images analysis showed alterations of mucosal pattern of ileal loops with an incresed number of folds (5 or more folds per inch) (“ileal jejunalization”) in 18 patients; reversal jejuno-ileal fold pattern in 6 patients; intestinal intussusception was observed in two patients and hyposplenism in one patient. MR was also able to identify mesenteric lynfadenopaty. No alteration of small bowel loops was observed in five patients. Conclusions: PEG-MRI is a feasible and reproducible imaging technique in both adult and paediatric populations; it may suggest a diagnosis of celiac disease as well as it is able to identify potential intestinal complications together with extra-intestinal findings.

S46

Abdomen and Pelvis: Imaging Conclusions: Rectal distention by water filling amy improve the depiction of primary rectal tumor and assessment of the perirectal tumor extension, but it does not significantly improve the accuracy for the regional lymph node involvement.

90 Staging of rectal cancer with MRI using phased array coil: Radiologic-Pathologic correlation F. Iafrate1, A. Laghi1, P. Paolantonio1, I. Sansoni1, R. Ferrari1, D. Marin1, M. Celestre1, M. Ferri2, R. Passariello1; 1Department of Radiological Sciences, University of Rome, Rome, ITALY, 2Department of Surgery, University of Rome, Rome, ITALY.

89 Preoperative local staging of rectal cancer: MR imaging with and without rectal distention M. Kim1, J. Lim1, Y. Oh1, J. Kim1, N. Kim2, W. Kim3; 1Diagnostic Radiology, Yonsei University College of Medicine, Seoul, REPUBLIC OF KOREA, 2Surgery, Yonsei University College of Medicine, Seoul, REPUBLIC OF KOREA, 3Internal Medicine, Yonsei University College of Medicine, Seoul, REPUBLIC OF KOREA. Purpose: To determine the usefulness of a rectal distention by intraluminally administered water filling for the depiction of the tumor and for the local staging of the rectal carcinomas on magnectic resonance (MR) imaging. Materials and Methods: Sixty-two patients with surgically proven rectal carcinomas underwent pelvic MR imaging before and after rectal distention by warm water. Four blinded readers reviewed each set of T1-weighted and T2-weighted axial images obtained before and after rectal distention, then recorded scores depicting the tumor. Receiver operating characteristic curve analysis was performed to determine differences in the determination of; 1) tumor penetration of the outer wall of the rectum, and 2) regional lymph node involvement. Pooled sensitivity, specificity, and accuracy of all readers were compared with McNemar test. Results: For all readers, tumor depiction scores were significantly higher in the distended images than in the non-distended images (p < 0.01). The overall sensitivity, specificity, and accuaracy for the perirectal tumor extention in the distended MR images (97%, 63%, and 87%, respectively) were significantly higher (p < 0.05, McNemar test) than those in the non-distended images (93%, 49%, and 80%, respectiviely). There was no significant difference between two image sets for the determination of the regional lymph node involvement.

Purpose: To assess the accuracy of MR imaging with phased-array coil in preoperative diagnosis and local staging of rectal carcinoma.To compare MR data, concerning T staging, sphincteral infiltration and lateral margin of resection, with histological specimens. To demonstrate pitfalls related to lesion staging. A correlation between pathological T stage and MRI data was reported in all cases. Methods and Materials: Thirty-two patients (22 males and 10 females; mean age: 64 y.o.) with known rectal cancer were studied using a 1.5T magnet (Magnetom Vision Plus; Siemens, Erlangen, Germany) equipped with phased-array coil. All patients had a biopsy proven tumour localized within 15 cm from the anal verge. MR protocols included a non breath hold high resolution T2 weighted Turbo Spin Echo (TR/TE/ETL/matrix7acq.t:4055 ms/132 ms/33/390x512/3 min 26s) sequence acquired on three axial planes. Diagnostic confirmation was obtained at surgery in 31/32 cases. One patients did not undergo surgery due to clear advanced T4 lesion and the diagnosis was confirmed on multiple imaging modalities (TRUS, CT, and MR). MR data, concerning T staging, sphincteral infiltration and lateral margin of resection, were compared with histological specimens in 31/32 cases. Results: On T2w images rectal wall layers were visible in all the cases; no differentiation between mucosa and submucosa was possible. Meso-rectal fascia was identified in all cases. At histology 10 lesions were staged as T1-T2, 22 as T3-T4. Of twenty-two tumours extending beyond the muscularis propria, eighteen were correctly identified and four (18%) were understimated. Overstaging did not occurred. Sensitivity of MR imaging in T staging was 81.8%, specificity was 100% and overall diagnostic accuracy was 87.5%. Positive predictive value was 100% and negative predictive value was 71.4%. The intraobserver agreement on tumour T stage was good (k=0.83,S.E.=0.12) Conclusion: The use of phased array coil provides good evaluation of rectal wall layers, spatial distribution of rectal cancer and an excellent visualization of the pelvic structures.In addition this technique is non invasive and useful for all rectal tumours irrespective of site and size. MRI was moreover highly accurate and reproducible in the staging of advanced rectal tumours and in predicting the lateral margin of resection with great accuracy and in identifying tumours involving adjacent organs, enabling correct selection of patients for preoperative radiotherapy, and avoiding the potential morbidity of unnecessary treatment in patients with a wide tumour free lateral resection margin.

Abdomen and Pelvis: Imaging

S47

Results: Prostate imaging with a voxel size of 0.57mm3 was successfully performed. Transverse slice reconstructions of decreasing slice thickness are shown in figure 1. It is evident that anatomical details are sharpened as slice thickness is reduced. The image contrast is T2-like.

Figure 1. 3D B-FFE transversal reconstructions of a) a 6mm slice, b) a 3mm slice and c) a 1mm slice. Images are cropped to 7 x 7cm2 field of view. Discussion: The use of an endorectal coil will increase the SNR especially in the posterior gland. Hence resolution may beimproved. The utilization of 3D radial k-space trajectories may remove problems with aliasing artifacts and allow shorter scan times. Conclusion: 3D B-FFE with nonselective block excitation pulses facilitates high-resolution prostate imaging with a T2-like contrast. It has been demonstrated that sub-millimeter resolution is feasible. References: 1. Storås T, Gjesdal K I, Geitung J T. 3D Balanced FFE imaging of the prostate. MAGMA 2002;15 Suppl. 1:268.

92 91 High resolution low SAR 3D B-FFE imaging of the prostate T. Storaas, MSc, K. I. Gjesdal, PhD; AMFT/ Medical Physics, Ullevaal University Hospital, Oslo, NORWAY. Introduction: The obtainable resolution of 3D Balanced Fast Field Echo (B-FFE) for prostate cancer staging is limited by radio frequency (RF) heating of the patient (1). We suggest a simple modification to the vendor implemented 3D B-FFE sequence that reduce patient heating significantly. The vendor implemented 3D B-FFE sequence utilizes a selective excitation pulse in order to avoid aliasing artefacts in the slice direction. This pulse is RF-intensive. Applying coronal slices, and imaging the entire anterior-posterior patient width remove the need for aliasing suppression in the slice direction. In this case B-FFE may be implemented with a less energy consuming non-selective (block) RF pulse. The purpose of this study was to investigate the feasibility of a 3D B-FFE sequence for high-resolution prostate imaging. Methods: The MR examination was performed on a Philips Gyroscan-NT Intera (1.5T) system with Power Track 6000 gradients. A two element flexible surface coil was applied. Pulse sequence modifications were done using a vendor-supplied interface. A coronal 3D B-FFE scan was prepared with the following parameters; repetition time = 5.0 ms, echo time = 2.1 ms, field of view = 148 x 148 mm2, scan-matrix = 160 x 240, number of slices = 350, slice thickness = 0.5/1.0 mm over-contiguous, number of samplings = 2. Scan time was 8 minutes and 58 s. The selective RF-excitation pulse was substituted with a block pulse of 0.5ms duration and 90° flip-angle. Selection gradients and corresponding rewind gradients were removed.

BOLD MRI of prostate cancer: Measurement of oxygenation changes during carbogen breathing P. Martirosian1, T. Diergaten2, F. Schick1, T. Wurm3, J. Hennenlotter3, A. Stenzl3, C. D. Claussen2, H. Schlemmer2; 1Department of Diagnostic Radiology, Section on Experimental Radiology, University of Tübingen, Tübingen, GERMANY, 2Department of Diagnostic Radiology, University of Tübingen, Tübingen, GERMANY, 3Department of Urology, University of Tübingen, Tübingen, GERMANY. Introduction: High-resolution endorectal MRI is an accurate method for local staging of prostate cancer, particularly for predicting extracapsular or seminal vesicle tumor invasion. However, improved imaging strategies are necessary to get information about functional tumor characteristics. It has been shown recently that T2*-weighted MRI provides a non-invasive method of assessing the response of individual tumors to Carbogen. Goal of this study was to establish an examination protocol for assessing oxygenation changes of normal and cancer tissue in the prostate during Carbogen breathing. Subjects and Methods: In this ongoing study 12 consecutive patients with biopsy proven prostate cancer underwent endorectal MRI at 1.5 T (Siemens, Erlangen, Germany). Studies of normal prostate were performed in 5 healthy volunteers using the pelvic phased-array coil. BOLD MR images were continuously acquired during four 4-min episodes alternating between air and carbogen (95% O2, 5% CO2) breathing. In each episode, forty images were acquired, using a multi-slice T2*-weighted echo planar imaging (EPI) sequence with TR=3000 ms, TE=60 ms, flip angle=90°, BW=1302 Hz/pixel, NA=2, in-plane resolution=1.56×1.56 mm2, and 12-14 slices of 3 mm thickness covering the whole prostate. Signal enhancement during carbogen breathing was quantified

S48

Abdomen and Pelvis: Imaging

from regions of interest after motion and baseline drift correction. The heterogeneity of tissue oxygenation changes could be visualized on activation parameter maps, which were calculated using a Student's t-test. Pixelwise correlation coefficients were thresholded at a level of 5.0 (p<0.01) for identification of pixels exhibiting significant response during Carbogen breathing. Results: In all examinations signal intensity enhancement of up to 20% could be detected in normal prostate tissue during Carbogen breathing (p<0.01). Signal changes in cancer prostate tissue were heterogeneous, but in most cases areas of cancer were characterized by lower signal enhancement compared to normal prostate tissue (Fig.1). The heterogeneity of tissue oxygenation changes could be visualized on the activation parameter maps.

Fig. 1. Axial T2-weighted MR image demonstrating extensive cancer in the right peripheral and central zone of the prostate. Signal enhancement during Carbogen breathing is higher in normal tissue (~8.8%) than in cancer (~5.5%). Conclusion: Multi-slice BOLD EPI sequences with high temporal resolution are feasible for detecting oxygenation changes in the whole prostate during carbogen breathing. Calculated parameter maps demonstrating the spatial heterogeneity of oxygenation changes in the prostate may be helpful for tumor detection and characterization.

MR Angiography and MR Contrast Agents I / Small Animal Imaging and Spectroscopy I Friday, September 19, 2003

Mini-Categorical Course MR Angiography and MR Contrast Agents I 8:00 am - 9:00 am

Willem Burger

93 Non-contrast MR angiography: overview of techniques (2D and 3D TOF, PC) G. E. R. C. Wilms, S. Sunaert, H. Bosmans, G. Marchal; Radiology, UZ-Gasthuisberg, Leuven, BELGIUM. The most frequently used and most robust technique for MRA of the intracranial vessels is the 3D-TOF sequence. With the appropriate technical measures to overcome saturation such as MOTSA, TONE and MT, excellent depiction of the intracranial vessels with high resolution distal anatomical detail is obtained. This high resolution is the major advantage over contrast-enhanced MRA. Therefore this technique can be used in daily clinical practice for the study of cerebral aneurysms and for the evaluation of vessel stenosis in atheromatosis and vasculitis. For the latter indications some overestimation of the degree of stenosis by turbulence has to be taken into account. PC (Phase-contrast)-MRA can be velocity-encoded as such that it simultaneously depicts slow and fast flow and therefore is the most appropriate technique to study cerebral AVM. Since PC implies subtraction of stationary background tissue, it is the method of choice to evaluate the underlying cause of cerebral parenchymatous or subarachnoid hemorrhage that superimposes on 3D-TOF images by the short T1 of fresh blood cloth. Venous anatomy and hence venous stenosis is best studied with PC-techniques providing the appropriate (slow) velocity encoding (VENC) is used. 2D-TOF techniques have few but very specific indications such as the differentiation between saturation induced signal loss and flowinduced phase effects and confirmation of venous thrombosis by slices perpendicular to the flow direction. Finally, for the study of the extracranial vessels non-contrast-enhanced techniques are obsolete, unless there is a clinical suspicion of dissection with fresh blood within the vessel wall. In the latter case, a T1w “black blood” technique can depict the fresh blood cloth as a linear intramural structure adjacent to the real lumen of the vessel, which appears completely black on this technique.

94 Contrast-enhanced MRA of the cerebral vasculature I. Berry, MD, PhD1, I. Catalaa, MD2, C. Cognard, MD2, C. Manelfe, MD2; 1Departments of Biophysics, Nuclear Medecine and Neuroradiology, University Hospitals Toulouse, Toulouse, FRANCE, 2Department of Neuroradiology, University Hospitals Toulouse, Toulouse, FRANCE. Contrast enhanced MR angiography of the cerebral vasculature takes advantage of the T1 shortening of the blood for preventing signal saturation. It also decreases the intravoxel phase dispersion occurring with turbulent flows. It depends strongly on the imaging parameters and on the time course after contrast injection. Appropriately set, with the use of rapid imaging, it provides series

S49

of dynamic MR digital subtraction angiography. This methods shows significant improvement in the assessment of brain arteriovenous malformations and aneurysms. With the dynamic series of acquisitions separating the arterial, capillary and venous phases of the contrast circulation, the nidus of AVMs is well depicted as well as the venous drainages, although still with limitation in temporal and spatial resolution as compared to catheter angiography. The evaluation of giant aneurysms benefits from contrast-enhanced 3D time of flight and dynamic contrast MR angiography because of less phase dispersion and release of T1 contamination artifact from blood clots. These methods allow for visualisation of aneurysmal sac and existing vessels better than the regular 3D time of flight MR angiography. Collateral flow and perfusion of compromised vessel territories are well assessed in the case of occlusions of the internal cerebral artery or middle cerebral artery.

Mini-Categorical Course Small Animal Imaging and Spectroscopy I 8:00 am - 9:00 am

Jurriaanse

95 State-of-the-art in generic manipulation of rodents B. Wieringa1, W. K. J. Renema2, R. in t Zandt2, A. de Groof1, E. Janssen1, A. Terzic3, P. Dzeja3, A. Heerschap2; 1Dept. Cell Biology NCMLS, University Medical Center Nijmegen, Nijmegen, NETHERLANDS, 2Exp.Radiology, University Medical Center, Nijmegen, NETHERLANDS, 3Div. Cardiomuscular Diseases, Dept. of Medicine, Molecular Pharmacology and Exp.Therapeutics, Mayo Clinic, Rochester, NY. MRS and MRI approaches are particularly well suited for in vivo study of metabolic homeostasis in healthy organisms or pathophysiological consequences of metabolic disease in animal models and human patients. In this area of science the study of (cellular) energy metabolism, mainly based on the use of 1H, 13C and 31PMRS for the monitoring of carbohydrate, adenylate or creatine (phospho) metabolites, has always been a central theme. With the advent of transgenesis, now almost two decades ago, entirely new avenues have been opened for the design of mouse models that either express mutant enzymes, or overproduce, or lack pertinent enzymes in pathways of interest. More recent developments even allow the design of “custom-build features” like conditional overexpression or conditional knock-out of particular enzymes. Also developmental-stage and cell-type dependent specification of expression of metabolic pathway enzymes is now possible. Our own work has been largely concentrated on the design and characterization of mouse models with defects in the integrated energetics network for high-energy phosphoryl (~P) transfer, including the Adenylate Kinase, Creatine kinase circuits, glycolysis or associated pathways. In these studies – which were mainly focused on muscle and brain - various aspects of major general importance became apparent: (i) Metabolic intervention (whether pharmacological or genetic) in energy metabolism, will always result in a global response, at the transcriptome, proteome and metabolome level. It is a general misconception, that these responses will only be found after developmental adaptation in conventional knock-outs or transgenic overproducers. Also in conditional mutant mice this will occur, predictably in any cell type in any

S50

MRI in Oncology: Diagnosis and Staging I

tissue. This demonstrates that there is a high degree of plasticity and redundancy in the cellular network for metabolic energetics, and global adaptations in the level and behaviour of enzymes and the changes in cyto-architectural properties are biologically significant. Changes apparently serve to compensate for the genetic defect and restructure the remaining system for optimal physiological performance under new conditions. Novel systems biology approaches are required to phenotype all changes. Aditionally, (ii) the often overlooked heterogeneous cell type composition of tissues and organs and the variation in genetic background of the transgenic animals is an issue of major concern, especially for brain studies, and deserves careful consideration before MRS or MRI can be applied for metabolic study. Acknowledgements:Research was supported by grants from the Netherlands Organisation for Fundamental Science (NWO-ZONMW and the Netherlands Cancer Society (NKB-KWF).

96 MR of animal models of human disease - strengths and weaknesses M. Rudin; CT/Analytical and Imaging Sciences, Novartis Institutes of Biomedical Research, Basel, SWITZERLAND. Biomedical research imaging is currently applied at various stages of drug discovery: Target validation, characterization of a disease phenotype, evaluation of drug efficacy and safety, and the development of imaging biomarkers that can be used for clinical drug evaluation. Focus area of structural and functional imaging have been down stream activities in drug discovery such as the assessment of drug effects in animal models of human disease. Imaging of the disease phenotype in animal models is the direct correlate to clinical diagnostic imaging. Today, MRI has evolved to the standard tool for morphological imaging in small animals due to high spatial resolution and excellent soft tissue contrast. There is a vast literature1 on using imaging and in particular MRI for pathomorphological characterization of animal models of human disease. Structural readouts comprise morphometric measures or parameters reflecting the tissue microstructure. Physiological parameters provide complementary information, which for many indications are very sensitive indicators of pathology. Examples comprise measurements of tissue hemodynamics, vascular permeability, tissue oxygenation, tissue function (brain activity, tracer clearance kidneys) to name a few. In the last couple of years cellular and molecular imaging approaches have been developed providing new insight into the pathophysiological cascade. In assessing drug efficacy, MRI either provides pharmacological endpoints (e.g. tumor volume) following therapeutic intervention or surrogates, i.e. measures that correlate with and, hence, predict a pharmacological endpoint1,2. The latter are of particular interest when investigating chronic diseases/disorders, for which clinical drug evaluation would be time consuming and expensive. Hence, biomarkers that give early indication on drug efficacy or safety would be of high value. Biomarkers are also of relevance for optimization of personalized therapy and for selection of patient populations for drug trials. It is obvious that biomarkers that are based on non-invasive technologies such as imaging will get a high acceptance by the clinicians. The most severe weaknesses of animal models are the models themselves, hardly ever reflecting full pathology as found in patients. Other limitations are the fact that studies have to be carried out in anaesthetized animals, an issue in particular for functional studies. With regard to molecular approaches, MRI suffers

from low inherent sensitivity, which together with the relatively bulky reporter groups that have to be used will preclude many potentially interesting studies. Today, the combination of complementary imaging modalities, with MRI being a major player, has become an indispensable tool for drug discovery and development 2.

Mini-Categorical Course MRI in Oncology: Diagnosis and Staging I 8:00 am - 9:00 am

Ruys

97 Paediatric CNS tumours C. A. Raybaud, MD; Neuroradiologie, CHU TIMONE, Marseille, FRANCE. Introduction: Frequency: most common solid tumors in children second only to leukaemia. More common in children than in adults. Variety: in contrast to adult, many various types (different glial types including ependymal, benign-malignant; neural, pineal; primitive neuroectodermal; germinal, other non neural; developmental neuroepithelial; developmental non-neuroepithelial; etc. Metastases exceptional, meningiomas usually related to NF2. Location: 50% posterior fossa; 30% 3rd ventricle; 20% hemispheres. Spinal cord tumors rare, mostly astrocytomas. Clinical picture: results from location: hydrocephalus; neurologic deficits (long tracts, cranial nerves, anterior optic pathways); endocronological disorders (growth defect, diabetes insipidus, delayed/precocious puberty); epilepsy not common. Diagnostic Tools: MRI: T1, T2, FLAIR, T1 (MT) with contrast (increases sensitivity, helps in tumor identification; useless for benign/malignant). Always 3 planes, spine pre-operatively for certain tumor types (medulloblastomas, germinal tumors mostly, but not only). Complementary: T2*, MT (without contrast), diffusion, SMR, perfusion, angio-MR, fMRI. Scanner: usually already done. Most useful without contrast: spontaneous density (germinal tumors, medulloblastoma), calcifications (pineal). Diagnostic Steps: Step 1: is there a lesion? Optimal efficacy of good MR investigation. But germinoma may become macroscopically apparent long after diabetes insipidus : repeat studies. Step 2: is the lesion a tumor? abscesses, other inflammatory (Schilder)aqueductal stenosis, benign cysts, cavernomas, etc. Step 3: is the tumor benign/malignant? edema, necrosis, hemorrage, invasiveness above all. SMR. Contrast enhancement not helpful in this respect for children. Step 4: where is the tumor? from the answer, both the diagnostic of nature and the surgical strategy will depend.Careful anatomy essential. Look pre-operatively for distant dissemination when needed. Step 5: effects on the brain? hydrocephalus, compression, infiltration, ischemia, herniations. Cisternal segment of cranial nerves important. Step 6: what tumor is it? use topography (most reliable), appearance and structure, mode of enhancement, evolutivity (duration of symptoms), invasiveness, age incidence (juvenile pilocytic astrocytoma, craniopharyngiomas, medulloblastoma, choroid plexus

3 Tesla and higher field strengths papilloma, etc.), clinical contexts (NF1, NF2, TSC), biological tumoral markers for germinal tumors. Conclusions: MRI necessary and usually sufficient for diagnosis, as well as for pre-operative setting. Necessary also for follow-ups (chemotherapeutic protocols). For brain tumors, the diagnostic rules used in adults do not apply in children. Many ways to approach the diagnosis, but one main triad: topography, appearance, clinical context.

98 Breast cancer C. Kuhl; Department of Radiology, University of Bonn, Bonn, GERMANY. Breast MRI is increasingly used as adjunct to conventional imaging modalities, particularly in diagnostic problem cases, but also for pre-operative staging. It is an extremely sensitive technique, with relatively limited specificity. To avoid both, false positive, but particularly false negative diagnoses, it is indispensable to be familiar with the varying MRI appearance of benign and malignant breast tumors. This course serves to give an overview on the different criteria that can be used for differential diagnostic purposes. Moreover, practical guidelines are given to help with the adequate management of enhancing lesions. Learning objectives: 1. To learn about diagnostic criteria related to breast lesion morphology, internal architecture, enhancement kinetics, and relaxation times. 2. To understand the role and the respective importance of these criteria for distinguishing benign lesions from invasive and intraductal malignant lesions. 3. To become familiar with the variable appearance of benign and malignant lesions in breast MR imaging. 4. To learn the pathophysiologic principles underlying the diagnostic criteria, and, accordingly, to understand the respective implications in terms of sensitivity and specificity of breast MR imaging 5. To learn about established and emerging indications for MR imaging of the breast

Plenary Session 3 Tesla and higher field strengths 9:20 am - 10:50 am

Willem Burger

99 Introduction of a 3T MRI system in the clinical practice: Do we have to reinvent the physics? P. Boesiger; University and ETH Zurich, Institute for Biomedical Engineering, Zurich, SWITZERLAND. During the last years, the main magnetic field strength of 1.5T has been the ‘gold standard’ for clinical MRI, covering virtually all MRI applications. Until recently, the use of higher fields has been confined to research, using long-bore systems not designed or approved for routine clinical use. Recently, however, ultra-compact 3.0T MRI systems came to market that offer the level of practicability required by the clinical environment, both in terms of installation and operation. The interest in high field systems is flourishing due several technical and clinical benefits of 3.0T field strength:

S51

·

Increased signal-to-noise ratio (SNR): Signal strength increases with the square of Bo, but the noise is proportional to Bo. Consequently, the signal-to-noise ratio (SNR) increases in a linear relationship to Bo, so that a doubling of the field strength also doubles the SNR. · Increased contrast: Increase of Bo changes the MR tissue contrast parameters: T1 is increased by about 30%, T2 and T2* are decreased. The changes in T1 are of particular benefit in MR angiography, resulting in a significantly improved vessel/tissue contrast. This may lead to an increase in vascular conspicuity at 3.0T. · Increased sensitivity: Functional MRI (fMRI) methods rely on the effect of blood oxygenation upon brain activation (BOLD effect). The change in relaxation rate R2* and subsequently the BOLD effect raises with Bo. · Increased chemical shift resolution: MR spectroscopy (MRS) benefits from increased chemical shift resolution. Improvements in spectral resolution enhance the possibility of spectral editing and speeding up spectroscopic imaging. · Parallel imaging: With higher resonance frequencies spatial coil sensitivity profiles become more distinct. Together with the higher SNR, higher field strengths basically allow for higher degree of parallelization. However for 3T the near field approximation dominates which clearly restricts this benefit. To understand these phenomena physics has not to be reinvented. The same formula apply, however with different numbers, leading to different appearance of the images and to different sensitivity for artifacts. To make use of the advantages and the benefits of higher field strength the examination protocols have to be adapted and optimized.

100 Body imaging applications at 3T S. M. Hussain; Radiology, Erasmus MC, Rotterdam, NETHERLANDS. Particularly, the development of a body coil and radio frequency coils for a 3 Tesla MRI system, and their recent approval in Europe and North America, facilitated its application as a whole body MR system. With about double the signal-to-noise ratio of a 1.5T system, the 3.0T MR systems dramatically improve the image quality and image acquisition speed. Higher signal-to-noise allows more imaging flexibility. Theoretically, 3.0T can deliver half the voxel size or, in certain cases, enable one-quarter of the acquisition time of a comparable study done at 1.5T. Additional resolution and speed could be used in a number of ways: 1) increased resolution: to visualize small structures in greater detail; 2) shorter acquisition times to reduce motion artifacts when imaging patients unable to perform extended breath holds; 3) improve both acquisition speed and resolution. Anatomy that is difficult to image at lower field strengths, such as the prostate, rectum, uterus, and ovaries, could be imaged without an internal coil, significantly reducing patient discomfort and preparation time. Faster breath-hold body imaging – with or without parallel imaging capability such as SENSE - is offered as standard feature by several vendors. In this lecture, we will present our experience concerning the body applications of 3T based on volunteer studies (n=10) and patients (n=50), and will explain some of the difficulties that occur during sequence optimization.

S52

Diffusion Imaging: Applications

Possible future body applications of 3T: • High resolution MRCP with heavily T2-weighted sequences • High resolution MR angiography (512x512 or higher matrices) • High resolution MR colography (possible comparison to CT if similar spatial resolution can be achieved at 3T, i.e. 512x512) • High resolution male and female pelvic MRI • High resolution musculoskeletal applications • Higher temporal and spatial resolution in liver imaging

101 High end neuro-imaging applications of 3T S. Sunaert; Dept. of Radiology, University Hospitals KULeuven, Leuven, BELGIUM.

indicated a recovery after the acute phase in RR-MS and a progressive shift towards abnormal values in SP-MS (for mean-values, the coefficient of the linear regression was r²=0.93). Thus, at the end of the study it was possible to differentiate the two subgroups while it was not the case at the onset. (the figures below show histograms for FA and Tr at the first and the final examination for RR-MS and SP-MS. For comparison, the average histogram of controls is also displayed). Discussion/Conclusion: These results confirm that DTI is sensitive to microscopic changes occurring in tissue of normal appearance on conventional imaging. Moreover, this technique allows to monitor these abnormalities over time. However, it is unable to clearly distinguish between the various physiopathological processes involved.

This talk will cover clinical neuro-imaging applications in several focussed areas – MR angiography, functional MRI, diffusion tensor imaging and spectroscopy – in addition to other clinical imaging applications which benefit from high field strength (3T). Advantages and disadvantages of high field versus field strengths less or equal to 1.5T will be discussed. We will also summarize specific technical improvements in hardware and software that are necessary to cope with some remaining disadvantages, such as increased susceptibility to artefacts and restricted use due to SAR limitations.

Scientific Session Diffusion Imaging: Applications 11:20 am - 1:00 pm

Willem Burger

102 Histograms of normal-appearing white matter of multiple sclerosis patients: a follow-up study with diffusion tensor imaging E. Cassol1, J. P. Ranjeva1, D. Ibarrola1, C. Manelfe2, M. Clanet3, I. Berry1; 1Biophysics and multimodality Imaging Neuroradiology MRI, University Hospitals, Toulouse, FRANCE, 2Neuroradiology MRI, University Hospitals, Toulouse, FRANCE, 3Neurology, University Hospitals, Toulouse, FRANCE. Purpose/Introduction: Our objectives were to determine the reproducibility of Diffusion Tensor Imaging (DTI) in volunteers and to evaluate the ability of the method to monitor longitudinal changes occurring in the normal appearing white matter (NAWM) of patients with Multiple Sclerosis (MS). Subjects and Methods: DTI was performed at 1.5 Tesla (Magnetom Vision, Siemens Erlangen, Germany). Six non-colinear directions of gradients were acquired with an EPI single-shot sequence (b-factor=506 s.mm-2). Six MS patients underwent threemonthly MRI examinations over one year: 3 had a Relapsing Remitting form of MS (RR-MS) and 3 a Secondary Progressive (SP-MS). They were selected with a limited cerebral lesion load. Age- and sex-matched controls underwent also three examinations. Diffusivity and anisotropy were quantified over the segmented whole supratentorial white matter, with the histograms of Trace (Tr) and Fractional Anisotropy (FA). Results: Results obtained in volunteers showed the reproducibility of the method. Patients had higher trace and lower anisotropy than matched controls (p<0.0001). Over the follow-up, both Tr and FA

103 Visualisation of tensor analysed q-space diffusion data S. Brockstedt1, J. Lätt2, R. Wirestam2, E. Larsson1, F. Ståhlberg2; 1Department of Radiology, Lund University, Lund, SWEDEN, 2Department of Radiation Physics, Lund University, Lund, SWEDEN. Introduction: High b-value diffusion imaging combined with tensor analysis allows extraction of information about fast and slow diffusion, and differences between white and grey matter have been shown [1]. The aim of this study was to obtain additional knowledge of extra- and intracellular water in cerebral tissue by applying tensor analysis to parameters obtained from q-space imaging. Subjects and Methods: Throughout the study a Siemens Magnetom 3.0T Allegra head scanner was used. A SE-based EPI sequence was modified to allow for diffusion encoding in six directions (dual gradients) using 32 equidistant q-values with qmax= 450cm-1 (bmax=6600 s/mm2), δ/∆= 20ms/88ms, Gmax= 38mT/m.

Diffusion Imaging: Applications Scanning parameters: TE/TR= 135ms/4000 ms, 10 slices, thickness 5mm, FOV= 215mm, 100×128 matrix. Five healthy volunteers (NEX=2) and five patients (NEX=1) were examined. Image analysis was performed using IDL (Research Systems, Inc.). Linear fits were made separately for two b-value regions (0-1200 and 22006600 s/mm2) and apparent diffusion coefficients (ADC) were determined for the two ranges, giving ADCfast and ADCslow, respectively [1]. From the displacement distribution obtained from q-space analysis, pixel-wise calculations of parametric maps showing the probability of zero displacement, FWHM [2] and the kurtosis [4] were made. Fractional anisotropy (FA) maps were also calculated based on tensor analysis of FWHM (32 q-values) and ADCfast (16 b-values) [3]. Directional anisotropy maps were obtained through colour coding of the largest eigenvalue. Results: Resulting images of a tumour patient are shown in Figure 1 where an increased ADC is seen in the edema surrounding the tumour. The kurtosis of the displacement probability function, obtained from q-space analysis, is close to zero in this region. This parameter might be used as a measure of the degree of compartmentalisation of different water components where low values imply free Gaussian diffusion of water. Figure 2 shows a comparison between conventionally determined FA maps and corresponding maps based on the FWHM value obtained after q-space analysis. Discussion: Tensor analysis of q-space data provides anisotropy maps of equal or improved quality as compared with conventional diffusion tensor imaging. Despite the fact that the improved quality of the FA maps partially can be explained by an increased number of measurements of the q-space data (twice as many), the additional information obtainable from q-space imaging makes further exploration of the method worthwhile. References: 1. Clark et al.[2002] MRM 47:623-628 2. Assaf et al.[2002] MRM 47:115-126 3. Basser et al.[2002] MRM 47:392-397 4. Lätt et al.[2003] 11th ISMRM

S53

104 Lipid droplet dynamics in BT4C rat glioma studied by 1H NMR diffusion spectroscopy in vivo T. J. Liimatainen1, P. Valonen1, M. Kettunen1, O. Gröhn1, S. Ylä-Herttuala1, R. A. Kauppinen2; 1Department of Biomedical NMR, A.I.Virtanen Institute for Molecular Science, Kuopio, FINLAND, 2School of Biological Sciences, University of Manchester, Manchester, UNITED KINGDOM. Introduction: It has recently been shown that 1H NMR spectroscopy (MRS) detectable lipids accumulate in rat glioma during gene therapy –induced programmed cell death (PCD) [1,2]. Substantial evidence supports the claim that 1H MRS visible lipids are in intracellular droplets of micrometer diameter. In this study the mean characteristic diameter (Φc) of the lipid droplets has been quantified by means of diffusion MRS during PCD. Subjects and Methods: Intracranial HSV-thymidine kinase positive BT4C gliomas were inoculated in the corpus callosum of rats and treated with ganciclovir for 8 days (1). Diffusion spectroscopy was performed at 4.7 T (VarianUNITYInova) using diffusion times (tD) between 98 and 538 ms, and diffusion gradient amplitudes G=0-16.8 G/cm applied in three dimensions (TE=38 ms, TR=2.0 s). With small q=(2π)-1γδG values signal attenuation as a function of q can be approximated by a linear model S(q)=S0[1-1/5 (πq)2Φc2] [3], which was fitted separately for each diffusion times. Results: Typical diffusion weighted spectra with different q-values are shown (Fig 1). Throughout the observation period, calculated average compartment diameters Φc, were independent of diffusion time within the range studied, indicating restricted diffusion. Interestingly, Φc, as calculated from 1.3 ppm resonance, had a nonsignificant (p=0.16) increasing trend during gene therapy (4.5±0.4 µm vs. 5.9±0.9 µm, n=5) with average slope 0.17±0.10 µm/day (Fig. 2). Characteristic diameter in untreated group between days 0 and 8 showed no change (5.9±0.5 vs. 6.2±0.5 µm, n=5) (Fig. 2). Variation in Φc between animals was greater than the PCD induced increasing trend. Conclusion: Our results indicate that the average lipid droplet size contributing to 1H NMR detectable signal is in order at 4.5-6.2 µm, consistent with values obtained recently with fluorescence microscopy [4]. Furthermore, a trend towards increase in diffusion space was detected during gene therapy–induced PCD. This may shed light to biochemical origin of the MRS detectable lipids.

Diffusion Imaging: Applications

S54 1. 2. 3. 4.

Nature Med. 5:1323-1327 (1999). Cancer Res. in press (2003). Magn. Reson. Med. 45:409-414 (2001). Cancer Res. 62:5672-5677 (2002).

Figure 1. Typical water suppressed 1H NMR spectrum of lipid resonances in BT4C glioma in vivo. q-value 0 (bottom) and 35773 1/m (top).

Figure 2. Mean displacement analysis results of 1.3 ppm resonance. ‘o’ average droplet size ± SEM (n=5) treated group, and ‘∆’ untreated group (n=5).

105 Diffusion MRI to assess antidotes to soman intoxication E. Kendall1, S. Eidt1, A. Obenaus2; 1Medical Imaging, University of Saskatchewan, Saskatoon, SK, CANADA, 2Radiobiology Program, Loma Linda University, Loma Linda, CA. Introduction: Soman (pinacolylmethylphosphonofluoridate) is a highly toxic organophosphorus cholinesterase inhibitor. Exposure causes hyperactivity of the cholinergic system (Shih, T-M and McDonough, J., 1997. Appl. Toxicology, 17(4):255). The ensuing status epilepticus (SE) contributes to profound neuropathology in the hippocampus, thalamus, amygdala and piriform cortex. We examined these neuro-degenerative sequelae using diffusionweighted magnetic resonance imaging (DWI) (Bhagat et al., 2001 Neurorep 12:1481). Here we report the non-invasive assessment of therapeutic intervention following administration of an experimental drug. GK-11 is a non-competitive NMDA antagonist that blocks cell activation induced by excitatory amino acids like glutamate. Methods: All protocols were approved by the University animal care committee. Male Sprague Dawley rats (group n= 6) given intra-peritoneal pretreatments of atropine sulphate (17 mg/kg) and bis-pyridinium oxime, HI-6 (125 mg/kg) 30 minutes apart were intoxicated with soman (180-200 mg/kg) administered subcutaneously 1 minute following the HI-6 treatment. Rats exhibiting robust seizures were imaged at 12, 24 48 hours and 7 days post

treatment on an SMIS 3.0T, 50 cm (130.6 MHz) scanner. A spin echo diffusion sequences (TR/TE 3200 ms/100 ms) using eight bvalues (0–30 000 s/cm2, d=17 ms, D=96 ms) normal to the plane of observation were used. Mean apparent diffusion values were computed from regression fit maps in bilateral ROIs (slice = –3.60 mm bregma). Diazepam (0.2 mg/kg; i.m.) and gacyclidine (GK-11, 1 mg/kg; i.v.) therapies were administered 1 min and 1.5 or 24h after soman. Results/Discussion: Hyperintense DWI signal was observed at 3 hours post treatment in all ROIs. At 12 h, this decreased in the retrosplenial cortex, hippocampus, thalamus, and amygdala below normal levels. The amygdala and piriform cortex exhibited a rise in signal intensities for up to 7 days after treatment. After 24h, the retrosplenial cortex, hippocampus and thalamus returned to normal DWI contrast. Diazepam or GK-11 therapy prevented soman induced ADC (acute ischemic type) changes. However, at 7 days post GK-11 (1 mg/kg; i.v.) therapy a decrease in ADC was observed in the piriform cortex (17% below normal). The data demonstrate that GK-11 provided neuroprotection against soman-induced neurotoxicity when administered 3 hours after soman exposure. This is the first report that provides temporal and spatial resolution using MRI of pharmacological interventions against soman-mediated, seizure-induced neuropathology. Summary: Soman-induced SE was accompanied by early changes within the hippocampus and the temporal lobe of the test animals. These changes sensitively detected by DWI.

106 MR diffusion anisotropy imaging investigation of neuroprotecting drugs on a rat spinal cord traumatic injury model A. Jasinski1, Z. Sulek1, T. Banasik1, D. Adamek2, P. Brzegowy3, K. Majcher1, A. Pilc4, T. Skórka1, W. P. Weglarz1; 1Radiospectroscopy and MRI, H. Niewodniczanski Institute of Nuclear Physics, Kraków, POLAND, 2Neuropathology, Jagiellonian University Medical College, Kraków, POLAND, 3Radiology, Jagiellonian University Medical College, Kraków, POLAND, 4Neurobiology, Institute of Pharmacology PAN, Kraków, POLAND. Purpose: To develop quantitative and noninvasive methods of assessment of neuroprotecting drugs limiting secondary excitotoxic injury after spinal cord trauma (SCT) using MR water diffusion anisotropy imaging (DAI) on a rat model of SCT. Subjects and Methods: 30 male Wister rats of 250 g to 300 g weight divided into 3 groups were used. A laminectomy at the Th13 spine level was performed and the SCT was induced using a dynamic weight-drop. MK-801 - an NMDA-receptor antagonist and MPEP - an mGluR metabotropic receptor antagonist were injected intraperitonealy immediately after the injury. MR DAI experiments were done at 4.7T with a Maran DRX console; using standard SE and modified FSE sequences with diffusion gradients applied parallel and perpendicular to the spinal cord. Dedicated inductively coupled probes were used to record MR im-

Diffusion Imaging: Applications ages of 128 x 128 with a FOV of 2 cm, slice thickness of 1.6 mm and gradient b-factors up to 1500 s/mm2. All experiments were ECG and breath triggered. Each rat was measured 4 times at 1h, 24h, 48h and 7d after the trauma. Data were analyzed using IDL based software developed in-house. Longitudinal diffusion DL = Dzz, transverse diffusion DT = (Dxx+Dyy)/2 and isotropy index ID = DT/DL were determined for selected regions in the white and gray matter of the spinal cord. Results: Good quality DW MR images, free from any motion artifacts were obtained from control and injured spinal cord of the rat in vivo. DW sagittal images delineate the traumatic region and its development in time very well as shown in Fig.1. Axial images taken through the center of injury, at 2.8 mm and at 5.6 mm show development of injury in different anatomical regions. Average values of ID for control rats after laminectomy are: IDWM = 0.2 ± 0.05 and IDGM = 0.5±0.1. After injury ID increases depending on the extent of damage. Application of MK-801 has little effect in the WM, but is clearly visible in the GM in the intermediate zone. Application of MPEP has effects in the WM and GM seen as a decrease of ID in the intermediate region. These DAI results are confirmed by subsequent histopathology. Conclusion: Diffusion anisotropy imaging requiring 1/2 time of full DTI experiment may be used successfully for testing neuroprotecting drugs on the rat spinal cord injury model. Acknowledgments: This work was supported by KBN, grant No 8 T11E 031 17.

107 Characterisation of mouse spinal cord injury in the acute phase by diffusion tensor imaging and multi-exponential T2 analysis J. Bonny1, M. Gaviria2, J. Donnat1, H. Haton2, A. Privat2, J. Renou1; 1Stim, INRA, Saint-Genes Champanelle, FRANCE, 2U583, INSERM, Montpellier, FRANCE. Purpose/Introduction: Our objective was to characterise an experimental spinal cord ischemic injury (SCI) of mice using ex vivo microscopy during the acute phase. It combines two complementary quantitative approaches ; diffusion tensor imaging (DTI) previously used in rats for assessing the different consequences of spinal cord trauma, and multi-exponential T2 analysis (MET2) known to inform about the anatomic distribution of the myelin. Subjects/Methods: The excised spinal blocks consisted in five vertebrae centred at the epicentre of the SCI. Experiments were performed at 37°C on an Avance DRX400 micro-imaging system. The study involves at least n=4 spinal blocks per delay post injury (+2h, +4h, +8h, +16h, +24h and controls). Images of five slices perpendicularly to the spinal cord were collected, each slice of 1mm thickness covering a single thoracic metamer (field of view of 5x5 mm2, matrix of 128x64). DTI and MET2 were performed by acquiring respectively 7 diffusion weighted images obtained in non-collinear directions and 16 spin echo images with equidistant TE ranging from 12.5ms to 87.6ms. After tensor estimation and discrete decomposition of T2 decays in each voxel, the values of

S55

the different parameters were measured by positioning 11 regions of interest in the spinal cord. Results: In WM, a rapid decrease of the largest diffusivity l1 parallel to the WM tracts is observed until 8h after SCI induction (t8h), whereas the radial diffusivities (l2 and l3) increase slightly. Afterwards, the behaviour of l1 depends on the location relatively to the focus of the lesion. Whereas a bi-exponential behaviour of the T2 decay curves is resolved in normal WM, the pathological process lead to an acute decrease of the amplitude of the short-T2 component at t8h. In GM, all the diffusivities tends to reach a minimum at t8h. The T2 decay curves exhibiting always a mono-exponential behaviour, the single relaxation time tends to increase towards a maximum at t8h. Discussion/Conclusion: Our results demonstrate that DTI and MET2 lead to quantitative markers which are both sensitive to the pathological events in the early period following the SCI. The changes in diffusivities are consistent with those reported in previous studies of rat SCI. The alteration of the amplitude of the shortT2 component in WM indicates a demyelination, especially straightforward at t8h. This characterisation of spatial and temporal evolution of SCI helps to define both therapeutic windows and optimal imaging parameters for an in vivo follow up of SCI in mouse.

108 Diffusion tensor imaging and fiber tracking of mouse skeletal muscle A. M. Heemskerk1, G. J. Strijkers1, M. R. Drost2, A. Vilanova i Bartroli1, K. Nicolay1; 1Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, NETHERLANDS, 2Department of Movement Sciences, Maastricht University, Maastricht, NETHERLANDS. Introduction: The mechanical properties of skeletal muscle are dictated largely by the type and spatial orientation of its constituting fibers. Knowledge about the local structure (e.g. fiber lengths and physiological cross-section) is therefore vital for the understanding of muscle function and adaptation. The goal of this study was to develop a tool for measuring the in vivo three-dimensional architecture of murine skeletal muscle, using diffusion-weighted MRI [1]. In combination with functional MR measurements of muscle contraction, this tool can be used to study muscle defects and muscle damage. Material and Methods: Specimen: Hindleg muscles from formalin fixated wild-type mice (n=6) were used for ex vivo experiments. Healthy Swiss mice (n=4) were studied in vivo, using 1.2-1.6% isoflurane at 0.4 l/min medical air flow for anesthesia. MRI: MR measurements were performed with a horizontal 9.5 cm bore, 270 MHz MRI scanner with a Varian imaging console, using a 1.5 cm solenoidal RF coil. A diffusion sensitized 3D turbo spin echo sequence was used with diffusion gradients applied in 6 directions and one reference image was recorded without diffusion weighting. Scan parameters were: FOV=15*15*30 mm3, matrix size=64*64*128, TE=8.5 ms, ∆=15 ms, δ=10 ms, b-value=0 or 835 s/mm2, turbo factor=6, NEX=2 and TR=1 s (total scan time 2:20). Fiber tracking: The fiber paths were calculated starting from a seed point in the direction of the principal eigenvector of the interpolated diffusion tensor matrix. Tracking was continued until the stop criteria (i.e. minimal fractional anisotropy and maximal allowed angle per step) were satisfied. Results: The mean values for the eigenvalues, the trace(ADC) and the fractional anisotropy (FA) as determined in the tibialis anterior

Musculoskeletal Spectroscopy

S56

muscle for the ex vivo and in vivo studies (Table) are comparable with previous studies in other muscle systems [2]. The figure shows the result of fiber tracking in the tibialis anterior (TA) muscle. Fiber tracking in the whole limb was able to reveal different muscle groups such as gastrocnemius and tibialis anterior.The fiber length in the TA (table) is in the same range as the optimal fiber length determined by dissection [3].

Conclusions: DTI for the in vivo measurement of fiber orientation in the mouse skeletal muscle was successfully implemented. This enables the study of muscle architecture in wild-type and transgenic mice. References: 1. Nicolay K. et al [2001] NMR in Biomed. 14:94-111 2. Damon B.M. et al [2002] MRM 48:97-104 3. Burkholder T.J. [1994] Journal of Morphology 221:177-190 λ1 ± SD λ2 ± SD λ3 ± SD Trace(ADC) ± (10-5 (10-5 (10-5 SD (10-5 cm2/s) cm2/s) cm2/s) cm2/s) in

1.72 ±

1.14 ±

0.90 ±

vivo

0.13

0.08

0.04

ex

1.31 ±

1.03 ±

0.85 ±

vivo

0.09

0.06

0.07

1.24 ± 0.08

Scientific Session Musculoskeletal Spectroscopy 11:20 am - 1:00 pm

Jurriaanse

FA ± SD fiber length ± SD (mm) 0.32 ±

6.3 ± 0.3

0.03 1.06 ± 0.06

and one with diffusion gradients off. Total scan time for diffusion images was approximately 3 minutes. Fiber tracking of skeletal myofibers was performed by determining the eigenvector associated with the principle eigenvalue. A spline function was then fit to the eigenvector so that points between pixels could be calculated. A fifth-order Runge-Kutta formula was used for the tracking algorithm, where stop constraints (such as low anisotropy and a large change in path direction) were employed [3]. After the tracking program was completed, the number of fiber bundles, fiber length and pennation angle were determined [4]. All calculations were performed using MATLAB (The Mathworks, Inc., Natick, MA). Results/Discussion: Preliminary work has determined 10 fiber bundles in the soleus during rest, with the pennation angle and fiber bundle length ranging from 11.2°±0.02° to 33.4°±0.12° and 0.36±0.03cm to 4.01±0.036cm, respectively. Data is presented as the mean ± standard deviation. The high end of these results are in good agreement with those found for pennation angle and fiber lengths in the soleus using muscle sonographs [5]. Fiber tracking by DT-MRI provides a promising method for determining fiber bundle orientation and length in skeletal muscle during rest and contraction. References: 1. Le Bihan D., et al., JMRI2001;13:534-46. 2. Damon B.M., et al., MRM 2002;48:97-104. 3. Basser P.J., et al., MRM 2000;44:625-32. 4. Ding Z., et al., MRM 2003;49:716-21. 5. Maganaris C.N., Acta Physiol Scand 2001;172:279-85.

0.22 ±

5.8 ± 2.4

0.04

109 Determination of muscle fiber orientation and length by fiber tracking using diffusion tensor images C. J. Galbán, S. Maderwald, E. R. Gizewski, H. H. Quick, M. E. Ladd; Department of Diagnostic and Interventional Radiology, University Hospital Essen, Essen, GERMANY. Introduction: Fiber tracking using diffusion tensor magnetic resonance imaging (DT-MRI) has been used successfully in visualizing white matter neuron bundles in the human brain [1]. This technique has been extended to skeletal myofibers, validated by more established techniques, and shown to reliably determine fiber geometry and orientation in vivo [2]. The purpose of this project is to demonstrate the use of fiber tracking in determining fiber orientation and length in the human soleus during rest and contraction. Methods: Using a 1.5 T Siemens Sonata, DT-MRI was performed on the calf muscle region of six healthy volunteers, during rest and while sustaining a plantar flexion contraction with a load of 4kg. A single-shot EPI sequence was used with a TR/TE of 2000ms/58ms. FOV, matrix size, slice thickness and number of slices were 24.9x18.7 cm2, 192x256, 3mm and 15, respectively. The diffusion tensor at each pixel was calculated using seven images for each slice: six diffusion-weighted images, with a b value of 250 s/mm2,

110 IMCL levels correlate with fiber orientation in human calf muscles. A 1H-MRSI study P. Vermathen, R. Kreis, C. Boesch; Dept. Clinical Research, AMSM, University and Inselspital, Bern, SWITZERLAND. Introduction: The concentration of intra-myocellular lipids (IMCL) is correlated with muscle fiber-type composition. Type I fibers contain three times more IMCL than type II fibers (1,2). In addition to fiber-type composition, fiber orientation also leads to variations in the way a muscle is used. This study aimed at correlating IMCL content in different calf muscles with fiber orientation, obtained from the same data using the orientation dependent dipolar coupling of Creatine and Taurine resonances. A correlation might indicate a direct relationship between IMCL and functional muscle properties. Methods: MRSI: Thirty measurements were performed on the calf of ten subjects (1.5T system, SIGNA, GE) using a 2D-MRSI sequence (transverse orientation, Matrix: 36x36, FOV: 22 or 20cm, slice thickness 15mm, TR=1600 or 1200ms, TE=35ms). Processing: After spatial zerofilling, apodization, and Lipid Extrapolation (3) the spectra were fitted using “TDFDFIT” (4) employing prior knowledge as previously described (5). The metabolites of each voxel were assigned to one of ten muscles using segmentation. Results: Average IMCL (area ratio to Cr in soleus) in different muscles are listed in Table 1. Significant IMCL differences be-

Musculoskeletal Spectroscopy tween some muscles were obtained with high values in soleus and 2-3 times lower values in tibialis anterior, tibialis posterior, or gastrocnemius muscles. Table 1 also lists muscle fiber orientations with respect to B0. A significant correlation between fiber orientation and IMCL content was found (R=0.79, p<0.0001, muscle as covariate, Fig. 1). Discussion: Qualitatively, the results are in agreement with morphometric studies: Soleus muscle comprises a high percentage (~88%) of type I fibers and, therefore, is expected to have high IMCL levels. In contrast, tibialis anterior comprises ~70% type I fibers and gastrocnemius, or extensor digitorum muscles only ~50%. However, these relatively small differences in fiber type composition can account only for parts of the IMCL differences between muscles. The significant correlation between IMCL and fiber orientation suggests that IMCL levels are also a consequence of different functional muscle properties, how the muscle is used and at which angle forces are applied. References: 1. Essen B [1977] Ann.N.Y.Acad.Sci. 301: 30-44 2. Howald H, Hoppeler H, Claassen H, Mathieu O, Straub R [1985] Pflugers Arch. 403: 369-376 3. Haupt CI, Schuff N, Weiner MW, Maudsley AA [1996] Magn Reson Med 35: 678-687 4. Slotboom J, Boesch C, Kreis R [1998] Magn Reson Med 39: 899-911 5. Vermathen, P., Kreis, R., and Boesch, C. [2001] Proc. ISMRM, 1871 Acknowledgement: Swiss National Foundation 31-053788.98

S57

111 Magnetic resonance spectroscopy shows an increase in intramyocellular lipid content in inactive muscle after exercise V. B. Schrauwen-Hinderling1,2,3, L. J. C. van Loon2, K. Nicolay3, R. Koopman2, W. H. M. Saris2, M. E. Kooi1; 1Radiology, University Hospital Maastricht, Maastricht, NETHERLANDS, 2Nutrim, Maastricht University, Maastricht, NETHERLANDS, 3Biomedical Engineering, Eindhoven University of Technology, Eindhoven, NETHERLANDS. Purpose/Introduction: Intramyocellular lipid (IMCL) content has been reported to decrease following prolonged submaximal exercise in active muscle and therefore seems to form an important local substrate source. Since exercise leads to a substantial increase in plasma free fatty acid (FFA) availability with a concomitant increase in FFA uptake by muscle tissue, we investigated the net change in IMCL content in both exercising and non-exercising skeletal muscle. Subjects and Methods: IMCL content was quantified by magnetic resonance spectroscopy in 8 trained subjects before and after a 3 hour cycling protocol (55% Wmax) in both exercising vastus lateralis and non-exercising biceps brachii muscle. Blood samples were taken before and after exercise to determine plasma FFA, glycerol and triglyceride concentrations. Subjects got a standardisation diet with a fixed macronutrient composition for three days before testing (15% of energy as protein, 55% as carbohydrate and 30% as fat) and were instructed to sustain from exercise. A MRS scan was performed before and after the 3 hour cycling protocol in the same voxel. The measurements were performed on a 1.5 T whole body scanner. The image guided localised 1H-MRS spectra from the regions of interest were acquired using a PRESS sequence (TR/TE 3000/25 ms, 128 averages, voxel size (12x11x18) mm3, and water suppression (CHESS)). The spectra have been fitted in the time domain using a nonlinear least-squares algorithm (AMARES (1)) in the jmrui software package (2) (http://www.mrui.uab.es/mrui/) using prior knowledge. Results: Whereas IMCL content decreased by 20.4±2.8% (p<0.001) in the m.vastus lateralis, we observed a substantial increase in IMCL content in the m.biceps brachii (37.9±9.7%; p<0.01). Both FFA and glycerol concentrations were substantially increased following exercise (85±6 vs 1450±55 µM; p<0.001 and 57±11 vs 474±54 µM; p<0.001, respectively), whereas plasma triglyceride concentration was decreased (1498±139 vs 703±57 µM, p<0.001). Discussion/Conclusion: We confirm earlier reports showing a decrease in IMCL content in active muscle following prolonged endurance exercise. The novel finding of our study, however is that IMCL content is substantially increased in non-exercising muscle. Therefore, we speculate that the elevated plasma FFA concentrations during the exercise trial resulted in an increased plasma FFA uptake in muscle, leading to a net increase in muscle triglyceride content in non-exercising muscle. References: 1. Vanhamme, L; van den Boogaart, A; Van Huffel, S; [1997] J Magn Reson 129: 35-43. 2. Naressi, A; Couturier, C; Devos, J.M; Janssen, M; Mangeat, C; de Beer, R; Graveron-Demilly, D; [2001] Magma 12; 141-152.

S58

Musculoskeletal Spectroscopy

112 Effects of aging on the IMCL content of slow-twitch and fasttwitch muscles in fa/fa and lean Zucker rats as measured by localized 1H-MRS M. Korach-Andre1, R. Deacon2, M. Beil1, D. Sun1, J. Gounarides1, J. Gao2, D. Laurent1; 1Core Technology, Novartis Institute for Biomedical Research, East Hanover, NJ, 2Diabetes Mellitus Therapeutic Area, Novartis Institute for Biomedical Research, East Hanover, NJ. Introduction: Intramyocellular lipids (IMCL) may be a primary cause of insulin resistance (IR). However, confounding animal data have shown that IMCL content decrease with age in spite of increased IR. Even more intriguing is the observation that physical training stimulates storage of fasting IMCL in thigh muscles, while improving insulin sensitivity. On the assumption that both age and muscle types are modulating factors of IR, this study explored IMCL time-course changes in slow-twitch and fast-twitch muscles of Zucker rats. Methods: Experiments were carried out on 5-week old male fa/fa (ZFR, n=6) and lean (ZLR, n=6) Zucker rats. Localized 1H-MR spectra were obtained on a Bruker 3.0T/60cm instrument using a 72mm birdcage resonator and a 3cm surface coil in a cross-coil fashion. 1H-NMR spectra were obtained from the tibialis anterior (TA) and soleus muscles of the left leg using a PRESS sequence (TR/TE=2s/18ms, 1200 averages) with water suppression. Peak areas for total creatine (tCr, internal reference), EMCL and IMCL were determined using a line fitting procedure. Data were collected every 10 days throughout the two-month study. Time-course changes in whole-body IR was assessed in a parallel study using the oral glucose tolerance test (OGTT). Results: At 5 weeks of age, ZFR exhibited a ~5-fold greater amount of IMCL in the soleus. Over the course of the next 3 weeks, the IMCL content in the TA muscle of ZFR steadily and markedly increased (p<0.05) almost by a factor 2, and then remained at a plateau value (IMCL/tCr ~3.8) until week 13. OGTT data showed an increased IR during the same 3 week period, plateauing over the subsequent 5 weeks. Interestingly, the IMCL content in the TA of ZLR, although subtly, steadily decreased throughout the timecourse of the study while no significant changes were measured in their soleus muscle. Discussion: These results suggest that IMCL may play a predominant role in fast-twitch muscles by triggering peripheral IR. The discrepancy between data obtained from the soleus and the TA muscles may be explained by the fact that slow-twitch muscles mostly rely on cellular lipid storage as a preferential substrate for contraction. In oxidative muscles, it is possible that fat cellular depots are compartmented within the muscle cell in such a way that glucose transport is not affected. In conclusion, proper timing should be considered to successfully test anti-diabetic drugs in the fa/fa Zucker rat.

113 Quantitfication of IMCL Signals and Peculiarities in the Spectral Pattern of 1H Single Voxel Spectra of Skeletal Muscle recorded at 3 T J. Machann1, C. Thamer2, M. Stumvoll2, F. Schick1; 1Department of Diagnostic Radiology, Section on Experimental Radiology, Tübingen, GERMANY, 2Department of Endocrinology, Metabolism and Pathobiochemistry, Eberhard-Karls-University, Tübingen, GERMANY.

Purpose: Quantitative assessment of intramyocellular lipids (IMCL) in human skeletal muscle by 1H MRS [1,2] is of increasing interest since it has been shown that IMCL are involved in pathogenesis of insulin resistance [3]. However, at a magnetic field strength of 1.5 T differentiation between IMCL and EMCL is aggravated due to overlapping signals mainly in muscle groups with oblique muscle fiber orientation. Aim of the study was to examine whether a examinations at 3 T lead to an improved differentiation between the mentioned lipid compartments. Spectral patterns were systematically compared for typical signal characteristics. Subjects and Methods: Examinations were performed on whole body imagers at 1.5 T (Magnetom Sonata, Siemens, Germany) and 3 T (Magnetom Trio, Siemens/Bruker, Germany) in the tibialis anterior muscle (TA) and the soleus muscle (SOL) applying a standard single voxel STEAM technique with TE=10ms, TR=2s, 40 acquisitions. Voxel size was chosen to 11x11x20 mm³ on both units. At 3 T additional spectra were recorded with smaller voxel sizes of 9x9x18 mm³ (48 Acq.), 7x7x14 mm³ (64 Acq.), and 5x5x10 mm³ (80 Acq.). Results: Besides the improved SNR (factor 1.7-1.8), IMCL and EMCL are clearly better distinguishable at 3 T in both muscles as the methylene resonance of IMCL shows smaller natural linewidths. Due to susceptibility effects, EMCL signals remain with a broad line shape in TA as well as in SOL. Cr3 and TMA signals are better resolved at 3 T in SOL, whereas in TA the TMA/Tau resonances are split in four signals at 3 T due to orientation dependent dipolar coupling effects [4]. Spectra from smaller voxels with careful positioning were shown to provide a clearly decreasing EMCL signal contamination, resulting in improved quantification of IMCL in SOL at 3 T. Discussion: In conclusion, muscular proton spectroscopy profits significant from higher field strength, as clearly narrower IMCL resonances and nearly unchanged susceptibility based broad EMCL signals occur. Reduction of voxel size is possible at 3 T because of the higher signal leading to a separate depiction of IMCL with reduced contamination by EMCL. Dipolar coupling effects were especially observed in TA (with parallel muscle fibre orientation to B0) and have to be further studied in detail. References: 1. Schick F, et al. [1993]Magn.Reson.Med. 29:158-167 2. Boesch C, et al. [1997] Magn.Reson.Med. 37:484-493 3. Jacob S, et al. [1999] Diabetes 48:1113-1119 4. Kreis R, et al. [1994] J.Magn.Reson. 104:189-192

114 Detection of changes in 13C-labeled phosphocreatine and creatine in human skeletal muscle after oral 13C-creatine intake M. van der Graaf1, D. W. J. Klomp1, M. Rijpkema1, J. J. A. van Asten1, M. Vlak2, G. W. Padberg2, A. Heerschap1; 1Radiology, University Medical Center Nijmegen, Nijmegen, NETHERLANDS, 2Neurology, University Medical Center Nijmegen, Nijmegen, NETHERLANDS. Introduction: Oral creatine supplementation (Cr-sup) is common practice in sports and medicine [e.g. 1], but no method has been available to monitor the dynamic uptake and clearance of the extra creatine. We used 13C-MRS in combination with oral intake of 13C-4-creatine for this purpose. Methods: One healthy male volunteer (53-yrs) on a low creatine diet ingested 20gr creatine/day for 5 days. 10% of the creatine was 13C-labeled at the guanidino carbon. Informed consent and approval of the local ethical committee was obtained.

Musculoskeletal Spectroscopy MRS measurements were performed on a 1.5T MR system (Magnetom Vision, Siemens Erlangen) using a 13C surface coil combined with a circular polarized 1H coil. 13C-MR spectra were obtained from the gastrocnemius muscle with an adiabatic pulse for excitation, and CW proton decoupling at 30W during the first 300ms of the data acquisition period of 819.2ms. Number of scans and repetition time (TR) varied: 900 scans with TR=1s for detection of the uptake of 13C-labeled creatine (Fig.1); 150 scans with TR=8s for the measurement of a fully relaxed spectrum (Fig.2); and 60 scans with TR=2s during an exercise (Fig.3) which consisted of a static force of approximately 600N applied at the ball of the foot. Also unlocalized 31P-MRS measurements were acquired with a 45° adiabatic excitation pulse, 48 scans and TR=12 sec, without NOE enhancement or proton decoupling.

Results and Discussion: As expected during the 5-day Cr-sup the PCr/ATP ratio (~ 4.4) increased by about 5% with respect to control [1]. Within 4 days of Cr-sup the 13C-creatine level reached a plateau of ~2x the original value (Fig.1). This is only slightly more than expected from the total creatine increase indicating little Cr turnover. After 3 months 13C-Cr was still elevated. Separate 13C signals were observed for PCr and for Cr (Fig 2) with Cr/TCr=0.28±0.01 (n=3). This is similar to the value found in mouse muscle (0.24) although in mice PCr/ATP is only 2.9 [2]. The assignments were confirmed by performing an exercise (Fig 3). Calculations of Cr/TCr from generaly accepted muscular ATP and/or total creatine values and PCr/ATP are lower than Cr/TCr obtained in this preliminary study. Conclusion: 13C-MRS can be used to monitor Cr uptake and wash-out, and to determine Cr/TCr in human tissue. References: 1. Vandenberghe K, Goris M, Van Hecke P, et al., J Appl Physiol. 1997;83:2055-63; 2. In 't Zandt H, de Groof A, Renema W, et al., J Physiol. 2003; 248.3:847-858.

S59

115 Failure of aerobic energy production throughout bacterial endotoxins infection in rat skeletal muscle. An in vivo study combining 31P-MRS and MRI B. Giannesini, M. Izquierdo, P. J. Cozzone, D. Bendahan; Centre de Résonance Magnétique Biologique et Médicale (CRMBM), UMR CNRS 6612, Faculté de Médecine de Marseille, Marseille, FRANCE. Introduction: Acute infections are associated with a well-recognized reduction of muscle performance but the exact mechanisms remain unknown. We have investigated the effect of bacterial endotoxin infection on rat muscle function using both 31P-MRS and MRI. Methods: Male Wistar rats (n = 10; 275-300 g) were injected i.p. with Klebsiella pneumoniae endotoxins (3mg.kg-1 body weigh) at times zero (t0) and 24h. Left gastrocnemius muscles were investigated before infection and at time 48h (t48). MR data were collected at 4.7 T in a horizontal magnet using a 31P-MRS surface coil (10x14 mm) and a 30-mm-diameter Helmholtz imaging coil, which are both integrated in a home-built imaging/spectroscopy probe. 31P-MRS acquisitions (TR = 2.4 s; 16 accumulations) were performed during 6 min of rest, 6 min of isometric contractions electrically induced at 3 Hz via surface electrodes, and 16 min of recovery. MR images (5-slices multiecho sequence; 8 echoes spaced 16 ms; TR = 1000 ms; acquisition time, 4.57 min) were recorded at rest and immediately after 16 min of recovery. Force was measured with a dedicated ergometer including a pedal and a pressure transducer. Absolute concentrations of phosphorylated compounds were expressed relative to [β-ATP] (5.8 mM) measured at rest. T2 changes, indicating muscle activation [1], were calculated on regions of interest chosen within the muscle section. The level of significance was set at P < 0.05. Results and Discussion: Endotoxin infection induces febrile hyperthermia, significant reductions of both body weight (-9.8±1.5 %) and of leg muscular volume (-10.2±3.6 %), and decrease in muscle performance. On the contrary, the pattern of T2 changes was not modified indicating that reduced muscle performance was not due to a failure of muscle activation. Infection did not affect the time courses of [PCr], [Pi] and [ATP] throughout the stimulation, but surprisingly induces both intracellular pH recovery and ADP accumulation during stimulation: at the end of the 6-min stimulation, pH reached 6.40±0.04 at t0 and 6.75±0.04 at t48, and [ADP] was 0.012±0.001 mM at t0 and 0.023±0.002 mM at t48. The analysis of PCr recovery kinetics in light of the different ADP concentrations measured at end of the stimulation disclosed a reduced oxidative capacity due to infection (19.8±2.2 mM.min-1 at t0, 13.2±1.5 mM.min-1 at t48). Conclusion: Bacterial endotoxin infection alters aerobic metabolism. This alteration could limit energy supply for contraction, hence reducing muscle performance. 1. Prior BM, Ploutz-Snyder LL, Cooper TG, Meyer RA [2001] J. Appl. Physiol. 90:615-623

S60

Musculoskeletal Spectroscopy

116 Robust homeostatic control of quadriceps pH during natural locomotor activity in man J. A. L. Jeneson1, F. L. Bruggeman2; 1Physiology, Division of Pathobiology, School of Veterinary Medicine, Utrecht University, Utrecht, NETHERLANDS, 2Molecular Cell Physiology, Biocentrum Amsterdam, Vrije Universiteit, Amsterdam, NETHERLANDS. Introduction: Classic invasive studies of quadriceps biochemistry during natural locomotor activity in man (i.e. two-legged (2L) exercise) have established acidification of at least 0.5 pH units during intense activity (1,2). Non-invasive 31P NMRS investigations have been restricted thus far to single leg (1L) exercise regimens (3). Here, we report on human quadriceps biochemistry during natural, 2L exercise. Subjects and Methods: Experimental. Six normally active subjects (5M/1F; age 24-36 yrs) were studied. Quadriceps pHi and cellular energy charge (CEC; [PCr]/ ([PCr]+[Pi]) at steady-states of ATP metabolism during incremental 2L-cycling were calculated from 31P NMR spectra (180 FIDs, TR 1000 msec; TA 60o) obtained at 1.5 T (4). Co-Response analysis (5). Regressions of pHi and CEC on scaled work rate (WR/WRmax) between rest and maximal WR for 1L (3) and 2L-cycling at 60 rpm were characterized by curve-fitting to compute partial-response coefficients RH+WR and RCECWR and the co-response coefficient ΩH+,CEC WR (=RH+WR/RCECWR) as function of WR. Results: Quadriceps CEC declined linearly (r2 0.81) from 0.92 + 0.01 at rest to 0.16 + 0.03 at maximal WR (mean 166 + 17 W; range: 108-223 W) translating to a depletion of 83 + 3% during incremental 2L cycling to exhaustion (Figure 1, lower panel). pHi fell linearly (r2 0.67) from 7.08 + 0.01 at rest to 6.84 + 0.02 (range: 6.93 – 6.78) at maximal WR, translating to a drop of only 0.24 + 0.03 units (Figure 1, upper panel). Co-Response analysis showed that homeostatic control of quadriceps pHi during 2L cycling, but not 1L cycling, is robust: the change in pHi per unit change in CEC was constant and moderate over the full span of steady-states of energy demand between rest and maximal 2L work (Figure 2). Conclusion: These results highlight the robustness of the Integrated System physicochemical and physiological control mechanisms in acid-base balance affecting local quadriceps biochemistry during natural locomotor activity in man. References: 1. Johnson RL, Heigenhauser GJF, Hsia CC, Jones NL & Wagner PD [1996] Handbook of Physiology. Section 12, 515-585. 2. Hermansen L & Osnes JB [1972] J Appl Physiol. 32, 304-308. 3. Richardson RS, Noyszewski EA, Kendrick KF, Leigh JS & Wagner PD [1995] J.Clin.Sci. 96, 1916-1926. 4. Rodenburg JB, de Boer RW, Jeneson JAL, van Echteld CJA & Bar PR [1994] J.Appl.Physiol. 77, 1021-1029. 5. Hofmeyer JHS & Cornish-Bowden A [1996] J.Theor.Biol. 182,371-380.

117 Phosphorylated guanidinoacetate is used for high energy phosphoryl transfer in creatine deficient knockout mice despite lower reaction rates H. E. Kan1, W. K. J. Renema1, D. W. J. Klomp1, D. Isbrandt2, A. Heerschap1; 1Radiology, UMC st Radboud, Nijmegen, NETHERLANDS, 2Center for Molecular Neurobiology, University of Hamburg, Hamburg, GERMANY. Introduction: The enzyme guanidinoacetate methyltransferase (GAMT) is a key enzyme in the biosynthesis of creatine (Cr), an important compound in the energy metabolism of skeletal muscle and brain. In the past decade, several patients suffering GAMT deficiency, and therefore lacking Cr, were described[1]. To enable in depth study of the disorder, GAMT deficient knockout mice (GAMT-/-) were generated[2]. In our study, we examined muscle energy metabolism of GAMT-/- mice during rest and ischemia using 31P magnetic resonance spectroscopy (MRS) and saturation transfer (ST). Methods: MRS measurements were performed on a 7.0 T magnet interfaced to a S.M.I.S. spectrometer operating at 121.53 MHz for 31P. A three-turn solenoid coil was used for 31P measurements, together with an Alderman-Grant type of proton coil for shimming. Mice were anaesthetized with 1,5 % Isoflurane and body temperature was maintained using a warm water blanket. For ischemia measurements in wild type (WT) and GAMT-/- mice (7 per group) a diaphragm plate which allowed reversible and reproducible obstruction of blood flow through the hind limb [3] was

Sequences and Techniques: Imaging used. Spectra were recorded with a pulse-acquire sequence (TR=1400 ms, 76 averages) for 7 minutes prior to ischemia, during the 25 minutes of ischemia and 16 minutes of recovery. ST measurements were performed on 3 WT and 3 GAMT-/- mice. The γATP signal was selectively saturated for several durations (500–5000 ms) prior to acquisition (TR=7000 ms, 64 averages). Data were processed using jMRUI [4]. Results: Prior to ischemia, the GAMT-/- mice showed a negligible phosphocreatine (PCr) content along with a new signal which was assigned to phosphorylated guanidinoacetate (PGua), the immediate precursor of Cr, as has also been observed in patients[1]. During ischemia, the decrease of the PGua signal was comparable to the decrease of PCr observed in wild type mice while its recovery after ischemia was slightly delayed (figure 1). ST measurements revealed a non-detectable flux from PGua to γ-ATP(figure 2).

Conclusion: These results show that PCr is absent in GAMT-/- animals. Interestingly, during ischemia guanidinoacetate is now used for high energy phosphoryl transfer at a rate comparable to PCr. However, since no flux from PGua to γ-ATP was detected during ST measurements, the corresponding enzyme kinetics are clearly reduced. References: 1. Schulze, A., et.al. Ann Neurol, 2003. 53 (2): 248-51. 2. Isbrandt, D., et al. J Inherit Metab Dis, 2002. 23: suppl 1:212. 3. in 't Zandt, H. J., et al. NMR Biomed, 1999. 12 (6): 327-34. 4. http://www.mrui.uab.es/mrui/mruiHomePage.html.

Scientific Session Sequences and Techniques: Imaging 11:20 am - 1:00 pm

S61

Ruys

118 Calculation of refocusing flip angles for CPMG-echo trains with predefined amplitudes: basic principles and applications for Hyperecho- and TRAPS-TSE J. K. Hennig, M. Weigel; Roentgendiagnostik, university hospital, Freiburg, GERMANY. Purpose/Introduction: The purpose of the paper is to present an algorithm based on the EPG-formalism (1,2), which allows to generate echo trains with predefined amplitude envelope. The algorithm is used to optimize low-SAR TSE-sequences like Hyperecho- and TRAPS-TSE-sequences(3,4) for high field applications. Subjects and Methods: All experiments were performed on a 3T whole body system (Siemens Magnetom Trio). Modified TSE-sequences with echo spacing esp=8ms and variable echo train length were used to verify the results of numerical simulations. In the extended phase graph algorithm magnetization is described as a superposition of phase states Fk and Zk, which can be shown to be related to the commonly used basis vectors Mx, My and Mz according to Eq.1 (s.below).

The effect of pulses and time evolution in between can be described by application of a sequence of transition matrices. The echo amplitude of the n-th echo is given by the population of state Fe0(n) at the time of each echo, which is related to the populations of states F1(n-1), F-1(n-1) and Z1(n-1) by Fe0(n)= E2 F-1(n-1)(cos(α(n)/2))2+ E2 F1(n-1)(sin(α(n)/2))2+E1 Z1(n-1)sin(α(n)) [2] E1 and E2 correspond to T1-, T2- relaxation terms respectively. The refocusing flip angle to achieve a given amplitude can be calculated by solving Eq.[2] for α. More generally, flip angles for N successive refocusing periods to achieve phase state configurations with specific populations of any number N of phase states Fk and Zk can be calculated. This is especially useful in order to prepare magnetization into the static pseudo steady state(SPSS) (6). Results: Fig.1 shows flip angles for an echo train with ETL=27 and a Gaussian envelope around the 13th echo. For initialization a 3ahead algorithm was used to prepare magnetization into the SPSS. The total RF-power of this sequence is 26.5% of that of a conventional TSE-sequence. Fig.2 shows a comparison of coronal images and images acquired with the sequence according to Fig.1 demonstrating the high image quality achieved. Discussion/Conclusion: The EPG-algorithm allows to produce echo trains with optimized amplitudes, which allow to perform high quality imaging at a fraction of the SAR of conventional TSE. 1. Hennig J. J.Mag.Res. 1988; 78: 397. 2. Hennig J. Conc.Magn.Reson. 1991; 3: 125. 3. Hennig J, Scheffler K, Magn Reson Med 46:6 (2001) 4. Hennig J, Weigel M, Scheffler K. Magn Reson Med. 2003;49:527-35 5. Alsop DC. Magn Reson Med. 1997; 37:176

S62

Sequences and Techniques: Imaging regions (indicated by arrows), as static. Multi-region NoQuist reconstruction used roughly 53 percent as much data showing faithful preservation of relevant image detail, with an adequate SNR for evaluation of myocardial functional parameters such as ejection fraction and radial thickening. Discussion/Conclusion: The Noquist method reduces acquisition time in dynamic MRI scans by eliminating the data redundancy associated with static regions. Our initial experience using Noquist with non-contiguous dynamic regions appears to yield stable and promising results. Non-contiguous dynamic regions in rFOV methods offer substantial further improvements in imaging efficiency, in particular in scenes with small and disjoint moving objects such as QF imaging of abdomen or neck.

119 Multiregion Noquist: non-contiguous dynamic regions within the field-of-view D. Moratal-Pérez1, M. E. Brummer2, J. Millet-Roig1, W. T. Dixon3; 1Ingeniería Electrónica, Universitat Politècnica de València, Valencia, SPAIN, 2Radiology Department, Emory University, Atlanta, GA, 3General Electric Global Research, General Electric Global Research, Niskayuna, NY. Introduction: Important dynamic MR imaging applications involve repeated acquisition of image planes in which only part of the field of view (FOV) changes over time. Conventional image acquisition techniques acquire all data for each image of such sequences. This represents a redundancy of the data. Several applications of cine MRI, importantly including cardiac MRI and quantitative flow (QF) methods, may likely benefit from faster acquisition. Two years ago we introduced a reduced-field of view (rFOV) technique “Noquist” for acceleration of ciné-imaging sequences. Here we report an important modification to this technique, demonstrating the technique’s capability of allowing multiple, non-contiguous dynamic regions within the FOV. Subjects and Methods: MRI data were acquired on a Philips Gyroscan Intera scanner using a balanced FFE sequence and a five element phased-array receiver coil. Typical parameters are: 192256 phase encodings, TR:3.5ms, TE:1.7ms, flip angle:60 degrees, FOV:250-350mm. Results: Figure 1 shows a short-axis slice through the heart of a normal volunteer. The multi-region approach allows classification of the heart and the great vessels as dynamic regions, reconstructing the rest of the FOV, including structures in-between dynamic

Figure 1. Full matrix(a), single-region “Noquist” reduced-matrix(b and c) and multi-region “Noquist” reduced-matrix(d) reconstructions from a short-axis study. In(a) and(d), an intensity profile along the white line in phase-encoding direction has been drawn. The arrows (b, c and d) point out the dynamic region considered within the FOV for “Noquist” reconstructions. In(c), the single dynamic region chosen for “Noquist” has been reduced in order to save more scan time than in(b), but the aorta, not the heart, causes a column of ghost artifacts(white box). In multi-region “Noquist”(d) this does not occur, because both, the heart and the aorta, are considered as dynamic regions within the FOV, saving more scan time than in (b).

120 Accelerating cardiac cine 3D SSFP imaging using k-t BLAST S. Kozerke1, J. Tsao1, R. Razavi2, P. Boesiger1; 1Institute for Biomedical Engineering, ETH and University Zurich, Zurich, SWITZERLAND, 2Guy's Hospital, King's College London, London, UNITED KINGDOM. Introduction: Cine 3D acquisitions of the whole heart within a single breath-hold have required compromises between image quality, spatial and temporal resolution. Using an elliptical k-space shutter, 3D data sets may be obtained at relatively high spatial resolution in a single breathhold (1). Still, temporal sampling does not satisfy the requirements for accurate volume determination (2). By exploiting spatio-temporal correlations in cardiac imaging, the acquisition efficiency can be considerably increased. The underlying method, called k-t BLAST, was introduced recently (3,4). The aim of the current work was to extend the k-t BLAST scheme to three spatial dimensions for cardiac-gated cine acquisition. To

Sequences and Techniques: Imaging improve the acquisition efficiency, sparse sampling and sampling of low-resolution training data may be obtained in a single scan. Methods: Sparse sampling of k-t space along a sheared grid pattern was implemented into a standard cine 3D SSFP sequence (Fig. 1). The acquisition of low-resolution training data, which were needed to resolve aliasing, was either interleaved into the sampling process or obtained in a separate prescan to shorten breathhold duration even further. Volumetric data sets covering the heart with 20 slices at a spatial resolution of 2x2x5 mm3 were recorded with 20 cardiac phases in a single breathhold of 20-23 sec, if training data were acquired in a separate prescan. Using interleaved acquisition of training data, breathhold duration was 25-28 sec. The feasibility of the method was demonstrated on healthy volunteers and on cardiac patients examined in a Philips Gyroscan Intera system.

Results: Volumetric data sets were successfully obtained in both healthy volunteers and cardiac patients at a temporal resolution of 35-45 ms. Taking the sampling of low resolution training data into account, the effective gain in acquisition efficiency was 4.8 compared to a standard, fully sampled 3D scan. Short-axis slices at endsystole obtained with 5x k-t BLAST in a patient are shown in Fig 2.

Discussion: Using k-t BLAST, cine 3D acquisition of the heart at 5-fold acceleration is demonstrated, making single breath-hold acquisitions at high spatial and temporal resolution possible. The reconstruction of nearly artefact-free images indicates the robustness of the approach, and functional parameters derived from accelerated acquisitions are comparable to those from standard single-slice data sets.

S63

References: 1. Jung BA, et al. MRM 2002; 48: 921-925. 2. Setser RM, et al. JMRI 2000; 12: 430-8. 3. Tsao J, et al. ESMRMB 2002; 45. 4. Tsao J, et al. ISMRM 2003; 209.

121 Continuous moving table acquisition using hyperHASTE U. A. Ludwig1, M. Weigel1, H. Fautz1, N. Ghanem2, O. Speck1, J. Hennig1; 1Department of Radiology, Section of Medical Physics, University Hospital, Freiburg, GERMANY, 2Department of Diagnostic Radiology, University Hospital, Freiburg, GERMANY. Introduction: Moving table acquisition techniques provide the advantage of continuous whole body imaging within a single measurement without temporal or spatial discontinuities of the acquisition. Maximum homogeneity in image intensity and contrast at a minimum of hardware and post processing requirements can be achieved with axial imaging (slice orientation perpendicular to motion). Thus, we established a T2- and/or STIR-weighted single shot RARE with hyperecho signal formation and Half Fourier reconstruction (hyperHASTE) in combination with a continuous table movement. Subjects and Methods: All experiments were performed on a 1.5T whole body scanner (Siemens Sonata) with use of local coil arrays. The moving table (angiosurf) was controlled by a RF shielded electrical drive. Due to SAR concerns, the RF power was minimized using TRAPS [1]. To reduce the spatial shifting of the refocusing pulses relative to the excitation pulse to 1/5*slice thickness, one of five widely spaced slices (distance 24mm) was acquired during each TR=500ms, while the table moved with constant speed (Fig.1). Therewith a five folded coverage of the entire body is achieved. Possible gaps between adjacent slices in one of the five axial sets due to respiration or other undefined motion are closed by the acquisitions of the other sets. One examination took about 12min (in-plane resolution=(1.6x1.6)mm2, ESP=4.6ms, TE=74ms, 300 repetitions). Alternatively, STIR-imaging was performed by means of an IR preparation (TI=160ms). Results: Fig.2 shows images of a) an original axial slice, b) a coronal reconstruction and c) a MIP reconstruction acquired with the STIR-hyperHASTE. The achieved quality of the original, axial images is very close to the one of conventional images. Neither breathing nor saturation artefacts due to the excitation of neighbouring slices are observed. The spatial shift in the slice positions of the phase encoding steps within the same echo train does not result in visible image artefacts either. Discussion: The employed single shot hyperHASTE proved to be motion insensitive to free breathing and table motion. No additional post processing and correction of field distortions was necessary for a gapless coverage of the human body. For faster table motions the frequencies of the refocusing pulses within one echo train can be synchronized with the table motion. The hyperHASTE sequence in connection with continuous table movement and local coil arrays allows a clinical protocol for fast screening and medical diagnosis with T2 and/or T1-weighted contrast of the whole patient’s body. [1] Hennig J., et al., MRM, 2003, 49, 527-35.

S64

Sequences and Techniques: Imaging The deformation is induced by a pneumatic piston, which applies a deformation of 1mm at the top of the phantom. Using a Siemens Symphony 1.5T whole-body scanner, the parameters for the modified STEAM-sequence were: TR=1270ms, FOV=140mm, 64×64 matrix, TM=150ms, TE1=22ms, TE2=26ms, Gd1×δ1= Gd2×δ2=100T/m/s in direction of the deformation (y) and in lateral direction (x). For comparison, two images using the standard STEAM-sequence with the same parameters (TE=22ms) and displacement encoding in y and xy direction were acquired. For a simple elasticity-contrast, the first spatial derivative of the displacement in encoding direction was calculated.

122 Detection of high-intensity focused ultrasound lesions using a new STEAM-sequence for static MR-elastography P. Siegler, L. R. Schad; Biophysics and Medical Radiation Physics, German Cancer Research Center (dkfz), Heidelberg, GERMANY. Introduction: In static magnetic resonance elastography (MRE) [1] the displacement inside a sample is measured between two static compression states. A new STEAM-sequence [2] allows a reduction of measurement time and number of deformations. Using this sequence, high-intensity focused ultrasound (HIFU) lesions were investigated, which cause decreased elasticity. Subjects and Methods: A STEAM-sequence with bipolar gradients for displacement encoding is normally used in static MRE. Replacing the second 90° pulse by two 45° pulses and switching a second bipolar gradient, it is possible to acquire two displacement components per excitation (see figure 1).

Fig. 1: Schematic of modified STEAM sequence. A HIFU-lesion resembling the number “1” (size: 10mm, line-thickness: 4mm) was induced in an ex-vivo porcine muscle (7×3×3cm3, acoustic power of 60W, 9s irradiation). To guarantee the recovery to its original state after compression, the sample was embedded into a cylindrical agarose-gel (diameter: 10cm, height: 8.5cm).

Fig. 2: HIFU-lesion (see arrows) inside porcine muscle: a) T2weighted image, b) photo through a cut of the investigated slice, c) and d) ∂uy/∂ry from the images calculated from the displacement images of the standard and modified STEAM-sequence, e) and f) ∂xy/∂rxy. Results: The first spatial derivatives of the displacement-data from both sequences look similar (see fig.2). The loss in SNR with the modified STEAM-sequence has no noticeable influence on the images. The position of the lesion can be recognised in ∂uy/∂ry and ∂uxy/∂rxy. Discussion: Using the modified STEAM-sequence it was possible to acquire two displacement components at the same time without adverse impact on the images. Using both displacement components together, the detection of HIFU-lesions might be improved. References: 1. Chenevert TL, et al. Elasticity Reconstructive Imaging by Means of Stimulated Echo MRI. Magn Reson Med 1998;39:482-490 2. Siegler P, et al. Stimulated Echo Sequence for Parallel Acquisition of Two Displacement Directions in Static MR Elastography. Proceedings of ISMRM 2003

123 Dynamic phonetic articulation assessment by gated MRI J. C. Vilanova1, J. Pujol1, J. Barceló1, S. Llach2, J. Cicres2, F. Pérez-Alvarez3; 1Magnetic Resonance, Clinica Girona, Girona, SPAIN, 2Department of Phonetics, University of Girona, Girona, SPAIN, 3Department of Pediatrics, Hospital de Girona, Girona, SPAIN. Purpose: To assess the phonetic articulation during repetitive speech using gated MRI, displaying 2D movies and to describe its applications on different fields. Methods: The phono-articulatory tract was imaged using a 1.5T MR unit, GE signa LX with gated SPGR (FOV 24 x24, matrix 256x128, TR: 30, TE: 3.2, BW: 15.6) employing a neurovascular coil. An external 1Hz trigger was fed to the cardiac gate. Subjects synchronized the sounds of phonemes to a flashing light synchronized with the external trigger. Each acquisition of 16 phases per second at a single-slice location was obtained in mid-sagital plane and took 2 minutes. All the vowels, selected consonants (stops, approximants, laterals, trills, fricatives and nasals) and three consonant vowel syllables (/apa/, /aka/, /ata/) corresponding to the sounds of the native catalan language were acquired. Results: The normal phonation structures were demonstrated for all the catalan vowels, consonants and consonant vowel syllables

Sequences and Techniques: Imaging sounds evaluated The single-slice data was displayed static and dynamically as cine loop for each acquisition. Low-pass filtering of data was helpful in removing noise and specially motion artifact from the trill consonants (r) . The relationship between the tongue, the palate, the lips, pharynx and epiglottis were well depict allowing a clear visualization either from the single image as from the cine loop images. Conclusion: Dynamic gated MRI of the phono-articulatory tract allows to visualize the movement of all its anatomical structures. This method provides a very useful tool in the phonetic field from every speaking language in various aspects of logopedics such as diagnosis and rehabilitation regarding the speech pathology and in evaluating defects in the velum and palate; but also in other fields like teaching phonetics and other languages, physiologic speech, research or design synthetized voices,

124 MRI of polymeric materials of biomedical interest: SPRITE imaging of hip prostheses P. Ramos Cabrer1, J. P. M. van Duynhoven2, A. van der Toorn1, K. Nicolay3; 1Experimental In vivo NMR, Image Sciences Institute, Utrecht, NETHERLANDS, 2Central Analytical Sciences, Unilever R&D, Vlaardingen, NETHERLANDS, 3Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, NETHERLANDS. Introduction: Success of hip arthroplasty and longevity of implanted elements largely depends on effective fixation of the prosthesis to bone surfaces (generally using cement). Conventional MRI techniques fail to provide good quality images of the status of such implants because of the nature of their constituent materials (low proton density and ultrashort T2* ranging 10-100µs) [1] and the presence of metal parts in the prostheses. This study shows the potential of the SPRITE technique [2] to image polymeric materials of biomedical interest, even in the presence of metals. High-resolution three-dimensional MR images of an acetabular socket from a hip prosthesis and the bone cement attached to it are presented. Methods: Polyethylene socket and attached (acrylic) cement used in the in vitro experiments come from a hip prosthesis replacement (~8 years implanted). Images were obtained using SPRITE in a 4.7 T horizontal bore magnet using a birdcage coil (6.4 cm i.d.). Experimental parameters used: 1.2µs excitation pulse (α = 4.5°); phase encoding time (~TE), 40µs; repetition time, 500µs; delay between two consecutive SPRITE ramps, 500ms; FOV, 6.4 x 16 x 6.4 cm3; data matrix, 64 x 160 x 64 (zerofilled to 128 x 320 x 128); acquisition time, 1h 30’. Results: A representative coronal slice and transversal slice as well as a 3D surface rendering reconstruction of the socket are presented in Figure 1. Polyethylene (white) is clearly distinguishable from acrylic cement (grey). Images show a gap between socket and cement surfaces demonstrating a non-ideal fixation. Polymeric parts of the RF coil are also visible, forcing us to acquire a large FOV to avoid aliasing artifacts on resulting images. In the images no artifacts originating from the metal ring in the object were observed. Discussion: It was shown that polymers of biomedical interest (like those used in prostheses) can be imaged in conventional MRI scanners using SPRITE techniques. Images are free from artifacts caused by B0 inhomogeneities, magnetic susceptibility gradients, or the presence of metal parts, etc. The extrapolation of these results to in vivo studies is challenging because of the large gradient amplitudes required in SPRITE techniques.

S65

Figure 1. Coronal and transversal views and a 3D reconstruction of a hip prosthesis socket. 1. Kennedy, CB et al., Can. J. Chem. 1988; 76:1753-1765 2. Balcom, BJ et al., JMR Ser. A 1996; 123:131-134

125 Design of highly selective 2D excitation pulses C. Barmet1, P. Boesiger1, W. J. Manning2, M. Stuber3; 1Institute of biomedical engineering, ETH Zürich, Zürich, SWITZERLAND, 2Cardiovascular Division, Department of Medicine, Department of Radiology, Beth Israel Deaconess Medical Center, Boston, MA, 3School of Medicine, Johns Hopkins University, Baltimore, MD. Introduction: In recent years the use of 2D selective excitation pulses was experiencing a renaissance: Respiratory navigator, spin labeling [1,2] and coronary vessel wall imaging [3] are examples. Some of these applications could benefit from an improved selectivity or a particular shape of the 2D pulse. Therefore a tool is needed to design the pulse, to study its shape and selectivity and to implement it on the scanner. The analytical solution of the Bloch equations (BE) for 2D pulses [4,5] (leaving two parameters undetermined) in combination with the numerical simulation of the BE is such a tool that is presented here. It was applied to design highly selective 2D pulses for arterial spin labeling [2]. Results of this work are the excellent agreement between simulated and measured pulse profiles as well as promising in-vivo-measurements. Methods: The analytical solution of the BE lets two parameters undetermined: the 2D k-space trajectory k(t) and the transverse magnetization Mxy of the pulse. Since the requirements on the pulse can be manifold (shape, selectivity, smooth plateau, aliasing pattern, duration), the choice of k(t) and Mxy is not obvious. The numerical solution of the Bloch equations for different parameter combinations will provide clarity. In arterial spin labeling [2] the blood in the ascending aorta is selectively labeled. Therefore a highly selective cylindrical 2D pulse is needed, enclosed by an aliasing free region of the size of the heart. For k(t) and Mxy the best performing candidates were found to be a spiral and a cylinder function respectively. Results: Highly selective 2D pulses were simulated, implemented on the scanner and measured in a phantom (fig. 1). In a volunteer, coronary arterial spin labeling was performed using these pulses (fig. 2). The labeled blood that enters the coronary arteries displays with high contrast. Conclusion: A tool was introduced that allows the design of 2D selective excitation pulses. Numerical simulations and phantom measurements show an excellent agreement. First in-vivo data have successfully been obtained. 1. Spuentrup E, Manning JW, Botnar RM, Kissinger KV, Stuber M [2002] Magn. Res. Med. 47:196-201. 2. Stuber M, Börnert P, Spuentrup E, Botnar RM, Manning WJ [2002] Magn. Res. Med. 47:322-329. 3. Kim YW, Stuber M, Börnert P, Kissinger KV, Manning WJ, Botnar RM [2002] Circulation 106: 296-299.

S66

Cells, Extracts, Fluids

4. Pauly J, Nishimura D, Macovski A [1989] J. Magn. Res. 81: 43-56. 5. Pauly J, Nishimura D, Macovski A [1989] J. Magn. Res. 82: 571-587.

Scientific Session Cells, Extracts, Fluids 11:20 am - 1:00 pm

ice, and dissected for medial thalamus and inferior colliculus (vulnerable brain regions) and frontal cortex (non-vulnerable region). Tissue samples were snap-frozen in liquid nitrogen, grounded over liquid nitrogen, and extracted with perchloric acid (PCA). Blood was taken from the neck and immediately mixed with PCA. The lyophilized samples were redissolved in deuterium oxide. 1H- and 13C-NMR spectra were recorded on a Bruker DRX-600. Results and Discussion: TD resulted in strongly impaired mitochondrial glucose metabolism in the neuronal compartments of medial thalamus and inferior colliculus, as evident from decreased TCA cycling ratios to 40% compared to pair-fed controls at presymptomatic stages, i.e. prior to the onset of neurological symptoms. In addition to a further decrease in TCA cycle flux at symptomatic stage, this was followed by a more than 55% decreased 13C-enrichment in acetyl-CoA entering the Krebs cycle and impaired flux through pyruvate dehydrogenase, resulting in decreased de novo synthesis of glutamate, GABA and aspartate. The symptomatic stage was further characterized by impaired glucose consumption and 4-fold increased lactate de novo synthesis. The onset of symptoms and region-selective impairments of energy metabolism correlated with decreased NAA at late stages (up to 45% of control). Comparison with [1,2-13C]acetate metabolism supported selective neuronal mitochondrial energy failure, while astrocytes showed increased glycolytic flux, pyruvate recycling, stimulated glutamine synthesis, as well as a limited support for neuronal energy- and neurotransmitter metabolism. In addition, astrocyte changes occured at very early stages (day 8), i.e. prior to the onset of behavioural symptoms and may contribute to deterioration of neuronal function during the progression of TD encephalopathy.

Plate

126 Investigation of the mechanisms contributing to neuronal energy failure in experimental thiamine deficiency: Role of astrocytic changes and intercellular trafficking C. Zwingmann1,2, D. Navarro1, R. Butterworth1, D. Leibfritz2; 1Neuroscience Research Unit, Hospital Saint-Luc (CHUM), Montreal, PQ, CANADA, 2Department of Organic Chemistry, University of Bremen, Bremen, GERMANY. Introduction: Pyrithiamine-induced TD in the rat reproduces a neurological syndrome resembling Wernicke`s encephalopathy. It has been proposed that decreased brain energy metabolism constitutes "the biochemical lesion" in TD encephalopathy. Since thiamine is a cofactor for several enzymes, ex vivo 1H- and 13C-NMR spectroscopy was used to investigate metabolic fluxes of [1- and U-13C]glucose as well as [1,2-13C]acetate (to recognize the astrocytic compartment) in thiamine-deficient rats. Methods: Rats were fed a thiamine-deficient diet and treated with the thiamine antagonist pyrithiamine. Presymptomatic animals were studied on days 11-12. A second group of animals progressed to acute symptomatic stages, characterized by loss of righting reflexes. Pair-fed control animals were placed on an identical diet, but administered with thiamine. After injection of 13C-labelled glucose or acetate, rats (n = 11-12 in each group) were killed by decapitation, the brains immediately frozen in isopentane over dry

127 Lactate: Preferred energy substrate for neurons? A. K. Bouzier-Sore1, P. Voisin1, P. Canioni1, P. Magistretti2, L. Pellerin2; 1Laboratoire RMSB, UMR 5536, Bordeaux, FRANCE, 2Institut de Physiologie, Lausanne, SWITZERLAND. Purpose: Nowadays, several data are in agreement with the idea that lactate could be a supplementary substrate for neurons, in addition of glucose, which is the major energy substrate for the brain. However, no information exists about the preferential use of either lactate or glucose by neurons when both substrates are available. Indeed, under physiological conditions, both glucose and lactate are present in the extracellular space, which result in a potential competition between these two substrates. We thus decided to study glucose and lactate competition on neuronal cultures to determine the relative contribution of each substrate to neuronal oxidative metabolism. Methods: Neuronal cultures were incubated for 4h in a medium

Cells, Extracts, Fluids containing either 5.5 mM glucose and 1.1, 5.5 or 11 mM [313C]lactate, or 5.5 mM [1-13C]glucose and 1.1, 5.5 or 11 mM lactate. 1H- and 13C-NMR spectra were realized on perchloric acid cell extracts (Bruker DPX 400). 13C-specific enrichments (SE) of different carbon positions of metabolites were obtained using a proton-observed carbon editing sequence. Results: Glutamate C4 SE rises with lactate concentration when lactate is the labeled substrate (from 22 % to 49 %) whereas it decreases when glucose is the labeled substrate (from 10 % to 3 %). Using the SE values of glutamate, lactate and glucose, we determined the relative contribution of each substrate to neuronal metabolism. The major labeled substrate observed in this study is glutamate. Since this metabolite is in equilibrium with the TCA cycle component α-ketoglutarate, the relative contribution of glucose and lactate to neuronal oxidative metabolism could be determined from the relative contribution of each substrate to glutamate synthesis. Under our experimental conditions three major pathways are involved in glutamate synthesis; from exogenous glucose, from exogenous lactate and from glycolytic lactate. A fourth pathway represents the endogenous glutamate isotopic dilution. When lactate concentration rises from 1.1 mM to 11 mM, then lactate participation to neuronal oxidative metabolism strongly increases from 47 % to 93 %. Conclusion: A major finding is that glutamate C4 SE is always higher when lactate is the labelled substrate. Calculation of the relative contribution of glucose and lactate to neuronal oxidative metabolism indicates that for a 1.1 mM lactate concentration and a 5.5 mM glucose concentration, glucose and lactate participates to the same extent to neuronal oxidative metabolism. When lactate concentration rises, then lactate participation strongly increases and represents 90 % of the main neuronal substrate.

128 Futyle cycling of lactate through the plasma membrane of C6 glioma cells as detected by 2H-13C NMR H. Gray1, T. B. Rodrigues1, S. Garrido1, R. J. Gillies2, S. Cerdan1; 1NMR Laboratory, Instituto Investigaciones Biomédicas, Madrid, SPAIN, 2Department of Biochemistry and Molecular Biology, Cancer Center, Tucson, AZ. Introduction: Tumor cells are known to metabolize the glucose provided by the neovascularization preferentially to lactate. Intracellular lactate produced by glycolysis is extruded to the extracellular space together with a proton using mainly the reversible monocarboxylate transporter MCT1. This process is thought to cause lactate accumulation and acidification of the extracellular space, two well known properties of the tumoral phenotype. However, because of the reversibility of MCT1 it is also possible that accumulated extracellular lactate re-enters the tumor cell establishing a futile cycle of lactate and protons through the plasma membrane. This report addresses this point and its potential consequences in pH homeostasis. Methods: C6 glioma cells were grown to confluence using DMEM (95%O2/5%CO2). On the day of the experiment, DMEM was substituted by Krebs Henseleit Buffer containing 50% (v/v) 2H2O, 2.5 mM (1-13C) glucose and 5 mM (U-13C3) lactate. Aliquots of the incubation medium were taken 0, 3, 6 and 9h after and analyzed by 13C NMR in a Bruker AVANCE 500WB spectrometer (125.13 MHz for 13C) using 1H decoupling only during the acquisition. Results: The Figure shows representative 13C NMR spectra of the lactate C3 region of the incubation medium sampled at 0h and 9h. It is possible to distinguish two different types of doublet reso-

S67

nances from the lactate C3 carbon. A doublet derived from the original (U-13C3) lactate and an isotopically shifted doublet derived from recycled (U-13C3, 2-2H) lactate. Thus, the presence of 2H2O in the incubation medium allows for the measurement of lactate recycling because the 2H incorporation in the C2 carbon through the intracellular lactate dehydrogenase equilibrium induces a vicinal isotopic shift of the C3 resonance, but only in the lactate molecules that have passed through the cytosolic space. After 9h of incubation approximately 50% of the extracellular lactate has been recycled through the plasma membrane. Discussion: The 13C-2H NMR approach described herein provides a convenient method to measure lactate recycling through the plasma membrane of cells. Our results indicate that this process occurs to a very significant extent. Lactate recycling produces and consumes the same amount of protons from the extracellular space and does not appear to affect extracellular pH. This may help to understand the lack of correlation observed between lactate resonances and extracellular pH in tumours in vivo (García-Martín et al.Cancer Res. 2001, 61, 6524-6531). Acknowledgements: Financed in part by Strategic group grat 2000-3 from the Community of Madrid.

129 Identification of microorganisms by NMR spectroscopy U. Himmelreich1, R. Somorjai2, B. Dolenko2, C. E. Mountford1, T. C. Sorrell3; 1Institute for Magnetic Resonance Research, University of Sydney, Sydney, AUSTRALIA, 2Institute for Biodiagnostics, National Research Council, Winnipeg, MB, CANADA, 3Westmead Hospital, University of Sydney, Sydney, AUSTRALIA. Introduction: Identification of microorganisms for clinical diagnosis is based on biochemical and assimilation tests, which take 2-3 days for some pathogens. Accuracy is in many cases compromised and safety of staff may be compromised by unanticipated exposure to infectious material. Delays in identification will delay implementation of appropriate treatment. More accurate molecular tests are available. However, these tests are also time consuming, laborious and expensive in their application. We propose a rapid, safe and accurate identification of microorganisms by NMR spectroscopy. Methods: Microorganisms: A total of 2700 clinical isolates of the following species were cultured on horse blood (bacteria) or Sabouraud dextrose agar plates: Yeast: Candida albicans, C.dubliniensis, C.glabrata, C.parapsilosis, C.tropicalis, Clavispora lusitaniae, Cryptococcus neoformans var. gattii and var. neoformans, Issatchenkia orientalis, Pichia guillerimondii and Saccharomyces cerevisiae. Bacteria: Acinetobacter complex, Citrobacter spp., Enterococcus

S68

Cells, Extracts, Fluids

faecalis, Escherichia coli, Klebsiella oxytoca, K.pneumonia, Morganella morganii, Proteus mirabilis, P.vulgaris, Staphylococcus aureus, S.epidermidis, Streptococcus agalactiae, S.milleri, S.pneumoniae, S.pyogenes. All yeast and bacteria isolates were identified biochemically. 10% of all isolates were also identified by PCR fingerprinting. NMR Spectroscopy: Colonies were suspended in 0.5 ml PBS/ D2O containing 107 to 5 x 109 cfu ml-1. 1H NMR spectra were obtained on a Bruker Avance 360 MHz NMR spectrometer using a 5-mm inverse-detection dual-frequency probe. 1H NMR spectra were acquired as follows: repetition time 2.3s, 4k data points, 32 transients, spectral width 3600 Hz, water suppression by selective excitation field gradients. Classification of NMR spectra: Comparison of two classification strategies was performed: 1. Selection of 16 integral regions using AMIX and classification by Linear Discriminant Analysis (LDA) according to [1]. 2. Using a genetic algorithm-based Optimal Region Selection (GA_ORS) for attribute reduction [2]. The averages of 2-4 most discriminatory subregions were used to develop pairwise LDA based classifiers. Results and Discussion: A simple LDA based classification of distantly related bacteria resulted in specificity and sensitivity between 90 to 98%. Less satisfying results were obtained with this method when applied to closely related species. After features reduction by GA_ORS, microorganisms were identified to the level of sub-species with an accuracy of 98%. Identification of microorganisms cultured in liquid media resulted in even higher accuracies. Incubation and preparation of liquid cultures can be fully automated if the NMR spectrometer is combined with a liquid sample preparation robot as known for drug screening applications. NMR spectroscopy hereby provids a save and rapid method for highly accurate identifification of pathogenic microorganisms.

130 Levels of activated sugars in breast cancer cells reflect metastatic potential as well as cell density. A 31P NMR spectroscopic study of cell extracts N. W. Lutz; Arizona Cancer Center, University of Arizona, Tucson, AZ. Introduction: Cancer metastases are responsible for the fatal effects of many malignant tumors. At present, it appears to be clear that cell-cell recognition plays a fundamental role in the invasiveness of cancer cells, and in their potential to attach to other tissues to form secondary tumors. Oligosaccharides in plasma membrane glycoproteins are important in cell-cell recognition and molecular targeting. This pilot study is aimed at gaining new insight into oligosaccharide metabolism in metastatic vs. non-metastatic breast cancer cell lines, and at exploring the potential of 31P NMR spectroscopy of activated sugars (= UDP-hexoses, UDP-Hex) to detect the metastatic potential of breast cancer. Materials and Methods: Sample preparation Two human breast carcinoma cell lines were used: MCF-7 (non-metastatic) and MDAmb-435 (highly metastatic). Cells were grown to 80% or 100% confluence, extracted with perchloric acid and prepared for 31P NMR spectroscopy as previously described [1]. NMR spectroscopy 1H-decoupled 31P NMR spectra were obtained at 202.5 MHz on a 11.7 T Bruker DRX500 NMR spectrometer using a broadband probe for 5-mm tubes. The acquisition time AQ

was 2.025 s, corresponding to 8k data points, and the sweep width SW was 2 kHz. Spectra were acquired at 4°C over 6-10 h with a repetition time of TR = 13 s (full relaxation). Statistics Two-way ANOVA was used to determine the significance of differences in relative UDP-Hex levels between groups. Results and Discussion: Two distinct UDP-Hex signal groups could be readily integrated, the β-phosphates of UDP-N-acetylglucosamine (UDPGlcNAc), and those of other UDP-Hex (oth.UDPHex) primarily consisting of UDP-N-acetylgalactosamine and UDP-glucose. Relative levels were determined by dividing UDP-Hexβ by NTPβ integrals (Fig. 1), a quantitation method commonly used in in-vivo 31P NMR spectroscopy [2]. All relative UDP-Hex levels varied as a function of both factors varied, i.e. metastatic potential and cell density (Fig. 1). This was confirmed by two-way ANOVA which also yielded signifcant interaction (p < 0.05) between these two factors for UDPGlcNAc and the sum of UDP-Hex (sumUDPHex). Increased UDPHex/NTP values for MCF-7 vs. MDAmb-435 cells have recently been reported for perfused cells embedded in matrigel [3]; however, our results clearly demonstrate that this difference becomes extremely small as cells reach high densities (close to 100% confluence). Fig. 1

References: 1. Lutz, NW et al. [1996] Magn.Reson.Med. 36:788-795. 2. Podo, F [1999] NMR Biomed. 12:413-439. 3. Sterin. M et al. [2001] Cancer Res. 61:7536-7543. Acknowledgement: Support by DOD grant DAMD17-01-1-0474 (to N.W.L.) is gratefully acknowledged.

131 Phosphocholine modulation by MAPK signalling M. Beloueche-Babari, M. O. Leach, S. M. Ronen; Cancer Research UK Clinical MR Research Group, Institute of Cancer Research and Royal Marsden NHS Trust, Surrey, UNITED KINGDOM. Introduction: Ras proteins are activated in many human cancers. The MAPK pathway constitutes a major Ras signalling component and inhibitors of this pathway are now in clinical trial. Using MRS to detect metabolic markers of MAPK signalling inhibition in cancer cells could, therefore, enable early non-invasive monitoring of response to this novel therapy. We used 31P MRS to investigate markers of MAPK signalling inhibition in human breast cancer cells and, since signalling inhibitors often act in a time- and dose-dependent manner, test whether these markers, if any, vary during the time course of the induced molecular response. Materials and Methods: MDA-MB-231 human breast cancer cells were treated for 16, 24, 32 and 40 h with 10 µM or 50 µM of the MAPK kinase inhibitor U0126 to achieve ~50% drop in cell number/flask at 40h. The effect of U0126 on P-MAPK levels was monitored using Western blotting. Cells were extracted at different time points following treatment using a dual phase extraction method and the aqueous fractions analysed by 31P MRS at 202 MHz. Results: Treatment with 10 µM U0126 induced a drop in P-MAPK levels at 16h which began to recover at 24h returning to control lev-

Cells, Extracts, Fluids els within 32h (Fig.1.A). 31P MR spectra of extracted cells indicated that phosphocholine (PC) levels dropped to 77±17% at 24h (p<0.01) but returned to control levels (107±18%, p=0.5) within 40h (Fig.1.B). No significant changes were observed in the other 31P MRS detectable metabolites. The transient effect on PC did not correlate with any changes in cell proliferation or cycling. To test whether this effect was related to MAPK signalling, cells were treated with 50 µM U0126 for the same durations. This produced total suppression of MAPK phosphorylation for up to 40h (Fig.1.C) and, interestingly, caused a sustained drop in PC levels equal to 69±10% on average (p ≤ 0.04) (Fig.1.D). Complete suppression of MAPK signalling in 2 other breast lines, namely MCF-7 and Hs578T, using 50 µM U0126 for 40h and 48h respectively, also produced a significant drop in PC levels. Conclusion: Our results show that MAPK signalling inhibition in MDA-MB-231 cells is associated with a drop in PC levels that occurs in a time- and dose-dependent manner. They also indicate that PC could be used as a non-invasive pharmacodynamic marker for monitoring inhibition of MAPK signalling when the blockade is sustained. Acknowledgement: Work funded by Cancer Research UK (grant C1060/A808)

S69

gelatin/Gd-DTPA solution. Four monolayers of 10 µl of cell suspensions containing 225, 550, 820 and 1100 cells, respectively, were added to each culture insert. High-resolution MR imaging was performed on the four cell concentrations using a General Electric 1.5T Signa CVi MRI using a 3D SPGR sequence with TR/TE=108.00/15.36 ms, flip angle = 60º, FOV = 1×1cm2, readout bandwidth of 7.81 KHz, Matrix = 512×384 and NEX = 8.0, scan time = 3 hours, slice thickness = 0.2 mm to 0.05 mm. A surface coil containing the sample fully was used for signal reception (inner diameter 1.5 cm). Results: Single myoblasts could be detected in all cell inserts, visualized as single voids from the magnetic susceptibility induced by the SPIO present in the phagocytic vacuoles. Conglomeration of multiple cells provided larger voids as a result of the higher concentration of iron per voxel. Good correlation was found between the signal void observed in MRI reconstructions and cell distribution under light microscopy. The volume of each cell suspension was estimated from 3D reformats using mIP, providing a volume similar to the 10 µl volume placed (11 µl ± 2.1 µl). The extent of the cell monolayer was approximately 1.2 mm. Discussion/Conclusions: Labeled cells were detected with resolutions down to 20×26×50µm3 voxels in vitro. SPIO labeled myoblasts will become feasible markers for the in vivo non-invasive monitoring of cardiac cell grafts in live animals and humans.

133 C-MLISA (Cellular Magnetic-Linked Immunosorbent Assay), a new application of cellular ELISA for MRI C. Burtea, S. Laurent, L. Vander Elst, R. N. Muller; Dpt Organic Chemistry/NMR Lab, University of Mons-Hainaut, Mons, BELGIUM.

132 Visualization of single myoblasts by high-resolution magnetic resonance imaging on a 1.5T clinical scanner P. A. Wielopolski, Ph.D.1, Z. Zhang, M.D., Ph.D1, E. van den Bos, M.D.2, G. P. Krestin, M.D, Ph.D.1; 1Radiology, Erasmus Medical Center, Rotterdam, NETHERLANDS, 2Experimental Cardiology - Thorax Centrum, Erasmus Medical Center, Rotterdam, NETHERLANDS. Purpose/Introduction: The introduction of myoblasts that can remodel damaged or dysfunctional myocardium to perform once again cardiac work is of great interest. Intracardiac transplantation of myoblasts has been previously shown to improve function. Myoblasts must be labeled with an intracellular probe to provide a high contrast mechanism that makes it possible to monitor them by MR imaging (MRI). Labeling can be performed with contrast agents such as Endorem (Guerbet S.A., Paris, France), a superparamagnetic iron oxide (SPIO) reagent. The purpose of our investigation is to demonstrate the detection of single pig myoblasts labeled with SPIO using high-resolution MRI on a 1.5T clinical scanner. Subjects/Methods: To facilitate SPIO uptake, Endorem was ligated with liposomes using lipofectamine 2000 (Gibco, USA) as our transfection reagent. Gelatin (3%, w/v) was added to a solution of Gd-DTPA (1:200 solution) in PBS, and 0.4 ml of this solution was added to a 1 cm diameter, plastic tissue culture insert. Fixed, labeled myoblasts were diluted in different concentrations in the

Introduction: The high throughput MRI has already been used [1] for the evaluation of the targeted nanoparticles. A modified cellular ELISA (C-MLISA) has been developed in the present work as an MRI application for clinical diagnosis. Two contrast agents with affinity for VLA-4 (very late antigen-4) integrin were synthesized by grafting the peptide GRGD and the CS1 fragment of Fibronectin (USPIO-g-GRGD, USPIO-g-FN) to USPIO particles. Both contrast agents were tested by C-MLISA on Jurkat cells and on rat mononuclear cells (MNC), which were stimulated to activate their integrins. Such an application presents interest for the in vitro detection of inflammatory pathologies, i.e. cancer, atherosclerosis. Subjects and Methods: Jurkat cells were stimulated with phorbol myristate acetate (PMA). Hepatitis was induced to Wistar rats with Concanavalin A, and the blood collected to isolate the MNC by Histopaque density gradient. The cells were suspended in buffered formalin and fixed on ELISA plates, which were subsequently blocked with a blocking buffer. The cells were incubated with the contrast agents (USPIO used as control) for 3 hours, and rinsed again with HBSS. The contrast agents bound to the cells were either digested with 5 N HCl, or resuspended by competition with the peptide. The specificity of the contrast agent for integrins was tested by pre-incubating the cells with the peptide. A calibration curve was constructed by incubating the cells with different concentrations of the contrast agent. The samples were analyzed by MRI (Bruker AVANCE-200, 4.7 T, TR/TE = 3000/20 ms, 20 echoes), and the T2 measured on images. Fe concentration was determined in each sample with Prussian blue, and the values correlated with T2. Results: The images (figure) show a striking difference between the stimulated cells incubated with USPIO-g-GRGD or USPIO-g-

S70

Neuro - Clinical

FN and the control ones. The significant correlation between T2 and [Fe] (r2 for cells +PMA = 0.987; r2 for cells -PMA = 0.998) demonstrate the specific interaction of the two contrast agents with integrins (i.e. VLA-4) (table). This specificity is confirmed by the pre-incubation of the cells with the peptide, which inhibits the binding of the contrast agent at the receptor sites.

Conclusion: C-MLISA shows promise as a new MRI method for the in vitro diagnosis of various pathologies (i.e. inflammatory), and could find application for the validation of specific contrast agents. References: 1. Högemann D, Ntziachristos V, Josephson L [2002] Bioconjugate Chem 13: 116-121.

Clinical Focus Session Neuro - Clinical 2:30 pm - 3:30 pm

Willem Burger

134 Motion processing in schizophrenia: a fMRI study H. Tost1, C. Wendt1, M. Ruf1, I. Wolf1, S. Brassen2, A. Schmitt1, D. F. Braus2; 1NMR Research in Psychiatry, Central Institute of Mental Health, Mannheim, GERMANY, 2NeuroImage Nord, Department of Psychiatry, University of Hamburg, Hamburg, GERMANY. Introduction: Eye-tracking dysfunction are frequently observed in schizophrenic patients and their biological relatives and a potential trait marker of the genetic disease vulnerability. Furthermore, recent laboratory data points to an underlying processing deficit in the dorsal visual processing network as a potential explanation for these oculomotor symptoms. In the present study, we investigate the hypothesis of poor smooth-pursuit eye movements in schizophrenic patients due to an impaired processing of target velocities in the human homologue of the motion-sensitive area MT (hMT, V5).

Method/Subjects: We investigated the brain activation of 12 stable medicated schizophrenic patients (mean age: 35.8 ± 7.2 years, 9 males, 3 females) and 13 matched controls performing a passive visual motion task. In 10 stimulation and 10 baseline blocks the presented visual paradigm comprised the pseudo-randomized presentation of tilted (120°) and moving (2°/s) sinusoidal gratings that alternated regularly with a fixation point. Imaging was performed on a commercial clinical 1.5 T MRI Scanner using a standard EPI sequence. Statistical procedures were carried out with BrainVoyager 2000. Results: In both groups, the combined visual stimulation provoked a robust activation of the primary and extrastriate visual cortices. Furthermore, hMT could easily be identified as a significant hot spot in the occipital temporal junction (BA 19) of both groups contrasting moving versus static condition (p<0.0001 corrected). Within the study groups, differences in activation pattern could only be identified concerning one circumscribed region located at the left junction of the cuneus and precuneus (BA 19/07) which was significantly more activated in controls while processing the moving gratings. The statistical comparison of the signal change and spatial extent of the individual hMT activation yielded no statistically significant group difference. Conclusions: In contrast to our hypothesis, we were not able to observe a functional abnormality of area hMT in schizophrenic patients while processing motion cues. This result argues against the explanation of the eye-tracking dysfunction and motion processing deficits of schizophrenic patients through a simple “bottom-up” visual processing deficit. Contrary to this, our results point to a deficient processing of motion cues at a higher level of the dorsal visual network usually associated with the top-down control of lower visual cortices. Further studies examining the functional correlates of active motion discrimination and smooth pursuit performance would further increase our understanding of the functional neuroanatomy of these symptoms (supported by DFG BR 1112/5-1).

135 Language lateralization in monozygotic twins discordant and concordant for schizophrenia: Evidence from functional MRI J. Tintera1, F. Spaniel2, P. Harantova1, M. Dezortova1, M. Hajek1; 1Department of Radiology, MR, IKEM, Prague, CZECH REPUBLIC, 2Psychiatric Center Prague, Charles University, Prague, CZECH REPUBLIC. Introduction: Studies reporting decreased language lateralization on the dichotic listening paradigm, decreased asymmetry of planum temporale and atypical leftward shift in the handedness distribution suggest decreased cerebral lateralization in schizophrenia. To asses relative contribution of genetic and non-genetic factors in etiopathology of schizophrenia, hemispheric dominance for language processing was examined using fMRI in 4 MZ twin pairs discordant and 4 MZ twin pairs concordant for schizophrenia. Results were compared with 4 healthy MZ twin pairs. Methods: fMRI measurements were done on Siemens Vision 1.5T with GE EPI sequence (TE=54 ms, TR=4 s, α=90°). Block design was used: 27 slices (THK=4 mm), 64 images, 8 images per each period of rest and stimulation. Subjects performed silent counting during rest and word generation during stimulations (verbal fluency task). Evaluation was done in SPM 99: Realignment, smoothing and normalization, square-box function (uncorrected p=0.001). Group of 4 MZ twin pairs discordant for schizophrenia with no familial loading of a psychotic illness, 4 MZ twin pairs concordant

Neuro - Clinical for schizophrenia with increased familial loading of schizophrenia and schizophrenia-spectrum disorder and four healthy MZ twin pairs were engaged in verbal fluency task. All subjects were righthanded. Activation was measured bilaterally in the frontal, temporal and temporo-parietal language areas, and a laterality index was derived from activity in these regions of interest in the left and the right hemispheres. Results: There was significant decrease in language lateralization in discordant subjects with schizophrenia compared to their healthy co-twins. Despite of their right-handedness concordant twins with high family loading for schizophrenia showed robust increase in variance of laterality index compared to discordant and control MZ twin groups. Only within concordant twin group right hemisphere dominance was found (three subjects). Conclusions: This finding suggests that genes controlling exact expression of cerebral dominance may be critically involved in etiopathology of schizophrenia. Acknowledgement: Study was supported by grant NF 67943/2001 of Internal Grant Agency of Czech Republic.

S71

volume enlargement or reduction in patients compared to control group. Results: In schizophrenic patients enlarged CSF spaces were found (p<0.001 corrected, cluster-level) . Volume reduction was found primarily in frontal (p<0.031 corrected, cluster-level), temporal (p<0.043 corrected, cluster-level) and cortical areas, in the thalamus (p<0.004 SVC) the cerebellum (p<0.011 SVC) and the amygdala (p<0.008 SVC).

136 Structural deviations in schizophrenic patients found with deformation-based morphometry W. Weber-Fahr1,2, P. Hubrich-Ungureanu1, G. Ende1, M. Ruf1, F. A. Henn1; 1NMR Research in Psychiatry, Central Institute of Mental Health, Mannheim, GERMANY, 2Department of Neurology, NeuroImage Nord, University of Hamburg, GERMANY. Introduction: Current research of schizophrenia emphasises a disruption in parallel distributed or dynamic circuits. Functional imaging studies have found abnormalities in the cortico-cerebellarthalamic-cortical circuit in different cognitive tasks [1]. Structural magnetic resonance imaging has convincingly demonstrated reduced volumes of the amygdala-hippocampal complex and other limbic and paralimbic structures [2], on both manual tracing and automated analyses. While several studies found local structural differences in GM and WM distribution through voxel based (VBM) or ROI-analysis morphometry, few attempts to identify larger scale differences through deformation-based morphometry (DFM) have been made. The purpose of this study was to corroborate the previous findings of structural abnormalities in schizophrenic patients with DFM. Subjects and Methods: 14 DSM-IV diagnosed schizophrenic patients were studied and compared to an age matched control group consisting of 13 healthy volunteers. Sagittal T1-weighted MRI scans were obtained with an isotropic voxel size of 1 mm³. Structural analysis of the anatomical data was done with deformation-based morphometry similar to the methods described in [3,4 ]. The anatomical data was co-registered with a standard T1-template in a two step normalization using SPM99. The first step includes a 12 parameter affine transformation to Talairach space, resizing to a voxel size of 2 mm³ and scalp-editing. The extracted brain images were then registered to the standard template using a nonlinear normalization procedure with a linear combination of 11x13x10 three-dimensional discrete cosine transform basis functions. As a result a deformation field for each subject with a specific threedimensional displacement vector in each voxel was obtained. By computing the Jacobian determinant of the displacement vector in each voxel a 3D-dataset for each subject was created containing information about local volume change. These data were tested in a voxel-wise T-test to identify regions of

Discussion: Deformation field based morphometry is able to fully automatically identify regions of altered volume in different brain regions of schizophrenic patients. In contrast to VBM, considerably larger structural region-changes of the brain shape can be identified. With this method, we found significant changes in brain regions previously reported to be abnormal in schizophrenia. References: 1. Andreasen Netal., Schizophr Bul 1998;24(2):203-18 2. Lawrie Smetal., Ann. N.Y. Acad. Sci. 2003;985:445-60 3. Ashburner Jetal., Human Brain Mapping 1998;6:348-57 4. Gaser Cetal., Neuroimage 1999;10,107-13

137 Multi-slice echo planar spectroscopic imaging in multiple sclerosis H. K. Mathiesen1, T. Tscherning2, P. S. Sørensen2, O. B. Paulson1,2, L. G. Hanson1; 1Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, DENMARK, 2Neurological Department, Copenhagen University Hospital, Rigshospitalet, DENMARK. Purpose/Introduction: Magnetic resonance spectroscopy provides information about neuronal loss/dysfunction by measuring loss of N-acetyl-aspartate (NAA). In multiple sclerosis (MS) most studies have been limited to single-voxels or measurements of whole-brain NAA (WBNAA) only. We present echo planar spectroscopy imaging (EPSI) as a promising alternative to non-localised spectroscopy for obtaining WBNAA estimates in addition to measurements of metabolites in specific areas (e.g. normal appearing white matter (NAWM), MS lesions). Subjects and Methods: The multi-slice EPSI sequence described

S72

Neuro - Clinical

in [1] was applied to patients with relapsing-remitting MS. Sequence parameters: TE/TR = 144/3360 ms. Eight 10 mm slices covered most of the cerebrum with 1 cm3 isotropic voxels. Matrix 32x32. Inversion recovery provided lipid signal nulling. Water suppression was obtained using a 32 ms chemical shift selective RF pulse. Conventional MRI was performed using FLAIR (fluid attenuated inversion recovery) on the same Siemens Vision 1.5 T wholebody scanner. Total scan time 40 minutes. The brain parenchyma suitable for spectroscopy was semi-automatically selected. Areas not suitable for evaluation (poor shim, CSF, short T2*) were excluded automatically. The brain mask was adapted manually to add consistency in the choice of regions and to remove residual areas known to degrade spectral quality (e.g. ventricles, frontal areas, eyes). Results: Spectra, WBNAA, and metabolites in specific areas were measured (Fig. 1 and 2). In this case, FLAIR was used to select ROIs, but other sequences can be used. Both NAA, choline, and creatine were measured. The latter is useful as a reference to correct for CSF content, coil sensitivity variation and oedema, since it is relatively constant. Discussion/Conclusion: Conventional MRI is essential in the diagnosis and follow-up of MS. However, the pathological specificity and correlation to disability is poor. MR spectroscopy correlates better to clinical disability, and has demonstrated pathological changes in NAWM. WBNAA has been estimated to decrease with age 10 times faster in MS patients than in the healthy controls, and has been suggested as a marker of disease progression and treatment efficacy. NAA loss in specific areas early in the disease evolution might have a prognostic value. However, studies with more patient are needed. We suggest the use of multi-slice EPSI, since this technique offers both WBNAA and metabolites from specific areas chosen after image acquisition. Acknowledgements: The project is supported financially by the Danish Multiple Sclerosis Society. References: 1. Hanson et al. [2000]. Mag Reson Med;44:412-417.

138 Monitoring Multiple Sclerosis on MRI: preliminary results of a multicenter comparison of early and late enhancement with 0.1 mmol/kg Gd-BOPTA versus early enhancement with 0.2 mmol/kg Gd-BOPTA A. Splendiani1, M. Gallucci1, L. Bozzao2, P. Pantano2, L. Regnicolo3, P. Cipolla4, A. Tartaro5; 1Servizio di Risonanza Magnetica, Università de L´Aquila, L'Aquila, ITALY, 2Servizio di Neuroradiologia, Policlinico Umberto I, Rome, ITALY, 3Servizio di Neuroradiologia, Ospedale Generale Regionale, Ancona, ITALY, 4World Wide Medical Affairs, Bracco Imaging SpA, Milan, ITALY, 5Dip. Scienze Radiologiche, Università G. D’Annunzio, Chieti, ITALY. Purpose: Contrast-enhanced MRI has high sensitivity for detecting blood brain barrier (BBB) defects inherent to acute inflammatory lesions of Multiple Sclerosis (MS). Diagnostic protocols for MR imaging of “active lesions” in cases of acute relapse typically envisage contrast agent doses of up to 0.3 mmol/kg bw. The aim of

Musculoskeletal Imaging this study was to compare the efficacy of single and double doses of Gd-BOPTA (MultiHance®, Bracco Imaging SpA, Italy) for the detection of lesions in patients with symptoms of new or relapsing MS. Methods and Materials: MR imaging was performed in 60 patients within 3 weeks of the onset of symptoms. PDw, FSE T2w, FLAIR and FSE T1w images were acquired pre-contrast followed by FSE T1w images at 5 and 15 min after a first injection of 0.1 mmol/kg bw Gd-BOPTA. A second injection of 0.1 mmol/kg bw Gd-BOPTA was then performed at 20 min after the first injection, followed, after 5 min, by the further acquisition of FSE T1w images. Preliminary on-site within patient comparisons were performed for the number and size of enhancing lesions on each post-contrast image set compared to pre-contrast. Safety assessments were performed before contrast administration, at 1 hour after administration and by interview at 24 hours after administration. Results: Although good visualization of lesions was noted on images acquired at 5 min after the first injection of 0.1 mmol/kg Gd-BOPTA, images acquired at 15 min post-contrast revealed a greater number of lesions and better lesion conspicuity. Images acquired at 5 min after the second 0.1 mmol/kg dose (cumulative 0.2 mmol/kg dose) did not reveal further improvement either in terms of lesion detection or lesion conspicuity. No serious or unexpected adverse events were reported. Conclusion: Imaging at 5 min after a cumulative dose of 0.2 mmol/kg Gd-BOPTA does not appear to provide any additional information beyond that obtained at 15 min after a single dose of 0.1 mmol/kg Gd-BOPTA for MR imaging of acute MS. The apparent requirement for only a single dose of Gd-BOPTA can be attributed to the high (9.7 mmol-1s-1) relaxivity of this agent compared to other gadolinium agents.

139 Is increased contrast agent relaxivity beneficial for enhanced MR imaging of brain tumors: blinded intraindividual comparison of Gd-BOPTA and Gd-DOTA C. Colosimo1, M. Knopp2, X. Barreau3, E. Gérardin4, M. Kirchin5, F. Guezenoc6, K. Lodemann7; 1Dip. Scienze Radiologiche, Università G. D'Annunzio, Chieti, ITALY, 2Dept. of Radiology, Ohio State University Hospital, Columbus, OH, 3Service de Neuroradiologie, CHU Pellegrin, Bordeaux, FRANCE, 4Service de Neuroradiologie, Hôpital Charles Nicolle, Rouen, FRANCE, 5World Wide Medical Affairs, Bracco Imaging SpA, Milan, ITALY, 6Altana Pharma, Altana Pharma, Paris, FRANCE, 7Bracco-Altana Pharma, Bracco-Altana Pharma, Konstanz, GERMANY. Purpose: To compare intraindividually a contrast agent with high T1 relaxivity (Gd-BOPTA, gadobenate dimeglumine, MultiHance; r1=9.7 mM-1s-1) with a contrast agent with conventional T1 relaxivity (Gd-DOTA, gadoterate meglumine, Dotarem; r1=4.3 mM-1s-1) for contrast-enhanced MRI of enhancing brain tumors. Method/Materials: 21 patients with suspected cerebral glioma or metastases received in blinded crossover fashion Gd-BOPTA and Gd-DOTA, each at 0.1 mmol/kg bw. Both agents were administered at 2 ml/s with an interval of ≥60h between exams. Images were acquired pre-dose (T1wSE, T2wFSE sequences) and postdose (sequential T1wSE sequences at 0,2,4,6,8,10,15 min with a T1wSE-MT sequence at 12 min) at 1.5T using a head coil. Each exam was standardized to ensure reproducibility. Assessment by two experienced fully blinded off-site readers was performed qual-

S73

itatively in terms of global contrast enhancement, global contrast agent preference, lesion delineation, lesion vascularization, and internal lesion morphology and structure, and quantitatively in terms of signal intensity enhancement and CNS-to-lesion contrast. Results: 19/21 patients were evaluated for efficacy. For the global assessment of contrast enhancement (primary criterion), reader 1 rated Gd-BOPTA superior in 18/19 cases (p<0.0001; Wilcoxon signed rank test) and the agents equal in 1/19 cases. Reader 2 rated Gd-BOPTA superior in 15/19 cases (p<0.001) and the agents equal in 4/19 cases. Consensus determined superiority for Gd-BOPTA in 18/19 cases and equality in 1 case (p<0.0001). Significant (p<0.0001) superiority for Gd-BOPTA was noted by both readers for all secondary qualitative criteria. Significant (p≤0.01) improvement in signal intensity enhancement and CNS-to-lesion contrast for Gd-BOPTA compared to Gd-DOTA was noted for both readers at all time-points from 2 min post-contrast. Conclusion: The two-fold greater relaxivity of Gd-BOPTA, deriving from a capacity for weak and transient interaction with serum albumin, permits improved contrast enhancement of enhancing brain tumors compared to Gd-DOTA. This may be important in patients with small or poorly-enhancing tumors which may be difficult to detect with conventional gadolinium agents.

Clinical Focus Session Musculoskeletal Imaging 2:30 pm - 3:30 pm

Jurriaanse

140 Additional value of magnetic resonance with spin echo (SE) T1 weighted imaging with fat suppression (FS) in characterization of soft tissue tumors J. L. M. A. Gielen, A. M. A. De Schepper, F. Vanhoenacker, P. M. Parizel, X. L. Wang; Radiology, University and University Hospital of Antwerp, Antwerp, BELGIUM. Purpose: The purpose of the study is to describe the signal intensity (SI) behavior of soft tissue tumors (STT) on spin echo (SE) T1 weighted images (WI) with fat suppression (FS) and to assess its additional value in tissue characterization. Subjects and Methods: MRI signal characteristics of 53 histological proven STT are discussed. Signal intensity behavior of STT could be classified in 4 types, representing specific tissues or tissue components. Type 1 is defined as low SI on both SE T1-WI and SE T1-WI with FS. Type 2 is defined as high SI on both sequences. Type 3 consists of high SI on T1-WI and low SI on T1-WI with FS. Type 4 is defined as SI comparable with SI of normal muscle on T1-WI and SI higher than normal muscle on T1-WI with FS. The additional information concerning contrast enhancement is described. Results: Type 1 SI behavior is noted in fibrous lesions, in hemosiderotic components, cysts and myxoma. Type 2 is noted in lesions containing methemoglobin or melanin. Type 3 is specific for fatty tissue. Type 4 is noted in highly cellular parts and in lesions of vascular origin. The use of SE T1-WI with FS improved lesion conspicuity on T1-WI. Discussion/Conclusion: SE T1-WI with FS has additional value in the characterisation of fibrous and hemosiderotic parts from cellular parts of lesions. It gives more confidence in characterisation of neurogenic tumors and hemangioma's. Presence of

S74

Musculoskeletal Imaging

methemoglobin and melanin are clearly discriminated from fatty tissue. Tumor conspicuity and inhomogeneity evaluation is improved. The use of SE T1-WI FS not only improves tumor conspicuity, but as tumor homogeneity and SI are important parameters in staging and characterization of STT, the use of SE T1-WI with FS will certainly be helpful. This may obviate the need for gadolinium administration

141 Grading and characterization of soft tissue tumors on Magnetic Resonance Imaging - A prospective study in 542 patients J. L. M. A. Gielen1, A. M. A. De Schepper1, F. Vanhoenacker1, P. M. Parizel1, J. Weyler2, X. L. Wang1, R. Sciot3; 1Radiology, University and University Hospital of Antwerp, Antwerp, BELGIUM, 2Epidemiology, University of Antwerp, Antwerp, BELGIUM, 3Pathology, University and University Hospitals of Louvain, Louvain, BELGIUM. Purpose: The purpose of our study is to evaluate prospectively the accuracy of MRI in grading and characterization of soft tissue tumors (STT) and pseudotumors. Subjects and Methods: The material consists of 542 untreated and pathologically proven STT examined by MR. A consensus report is made by two experienced radiologists using a subjective analysis of the MR imaging characteristics. A specific diagnosis or differential diagnosis with a maximum of three possibilities is obtained. Pathology results are confronted in a prospective way with the MR diagnosis. Concerning grading true negatives are benign lesions with benign MRI features, true positives are malignant lesions with at least one malignant MRI diagnosis given. Tumors are classified according to pathological phenotype. Results: 125 patients had a malignant STT, 417 a benign one. Concerning grading a specificity of 82%, PPV of 60%, sensitivity of 92%, NPV of 97% with accuracy of 84% is obtained. Concerning phenotype characterization on the first MRI diagnosis an overall sensitivity of 69%, specificity of 98%, accuracy of 96%, PPV of 71% and NPV of 98% is obtained. For benign lesions a sensitivity of 58%, specificity of 99%, PPV of 76%, NPV of 97% and accuracy of 96% is obtained. A match between the first proposed MRI diagnosis and pathology was found in 67%.The exact histology of all lesions was predicted in 61%. Conclusion: Our study shows that MRI reliably identifies malignancy in STT. False negative (n = 5 ) cases without biopsy proposition may go undiagnosed and untreated and their number has to be as low as reasonably achievable. Concerning phenotype classification of all lesions the best sensitivity and specificity was found in vascular, fatty and neural phenotypes. STT (benign and malignant) are confidently differentiated from pseudotumors. In malignant lesions the exact histology was predicted in 46%. On the contrary, in benign lesions the exact histological diagnosis was made in 88% of lesions. In conclusion, the results of our prospective study, are obviously better than those reported in previous studies. Good sensitivity is probably the result of progress in radiological science with inclusion of newly described parameters, description of larger series of tumors with specific imaging characteristics and an increasing experience. MRI further improves grading and characterization of STT and will

be together with proliferation markers, histological and histochemical examination a cornerstone method in diagnosis avoiding non suited invasive therapy in many benign and optimizing treatment, prognosis and outcome in all STT.

142 T2 relaxation time reduction in articular cartilage with in situ compression of the knee G. P. Liney1, D. Nag1, P. Gillespie2, K. P. Sherman2; 1Centre for MRI, Hull Royal Infirmary, Hull, UNITED KINGDOM, 2Department of Orthopaedics, Hull Royal Infirmary, Hull, UNITED KINGDOM. Introduction: Osteoarthritis (OA) represents a major cause of morbidity in the adult population. However, at the time of diagnosis, the disease has already advanced to an irreversible state. Detection of in vivo changes is critical in order to diagnose early OA. This study concentrates on the feasibility of measuring T2 to examine the in-vivo physiological changes in articular cartilage of the knee on MRI with in-situ axial loading in aymptomatic subjects over a wide age group. Methods: A total of 26 subjects aged between 27 and 73 years were studied. An MR-compatible mechanical loading jig was constructed and calibrated to deliver 60 kg of compression to the knee in situ. MR imaging was performed on a 1.5 T Philips Intera scanner. Anatomical images of the knee were acquired using a 3D fat suppressed FSPGR sequence (TE/TR/flip = 9/39 ms/35°) with 1.5 mm slice thickness. A T2 mapping protocol was used which consisted of 8 multiple spin-echo images (TE/TR = 20-160/2000 ms). Validation of accuracy and reproducibility was performed on one subject using gadolinium-doped agarose gel phantoms strapped to the knee and imaged over the course of one hour. T2 measurements were recorded in 6 zones of tibial and femoral cartilage as determined from the anatomical images. The post-load images were corrected for any movement and rotation by cross-referencing in two image planes. Results: Mean values of T2 in tibial cartilage were smaller than femoral cartilage, (38.0 ± 15.4 ms; 55.2 ± 6.2 ms, P < 0.001). Loading significantly reduced (P ≤ 0.001) T2 values in the most medial (1) and lateral (6) zones: 53.7 ± 8.7 ms vs. 46.6 ± 10.8 ms; and 59.3 ± 8.6 ms vs. 49.7 ± 12.0 ms. There were no significant changes in tibial cartilage. Young subjects (below 60 years), demonstrated a significant reduction in zone 1 of the femoral cartilage. Older subjects demonstrated significant reductions in zones 1 and 6. Figures 1 & 2 show plots of T2 values in the femoral cartilage versus age for each subject, with unloaded/loaded values indicated by open/filled squares and dashed/solid lines indicating respective fits to these data.

Musculoskeletal Imaging

S75

presence of two different water compartments in slow exchange with respect to T2 relaxation. The T2 values of these component are within the range of T2 found in in vivo animal studies [3,4]. Nevertheless, additional study is required to assign these components to the microanatomical compartments within the median nerve. 1. Beaulieu C et al. [1998]Magn.Reson.Imag. 16:1201-1210 2. Peled S et al. [1999]Magn.Reson.Med. 42:911-918 3. Does MD et al. [2000]Magn.Reson.Med. 43:837-84 4. Does MD et al. [2002]Magn.Reson.Med. 47:274-283 T2 relaxation times and amplitudes (A) of the relaxation components within in vivo nerves.

Conclusion: This study has demonstrated the feasibility of T2 quantification of articular cartilage with and without in situ loading. There are significant changes in T2 relaxation time, which may be due to variation in water content and may reflect the loadbearing status of cartilage.

I T2(ms) A(%) Mean ± STD Ref[3] Ref[4] 12 ± 0.1 19 ± 3

II T2(ms) 20 ± 4 41 ± 7 33 ± 4

A(%) 75 ± 8 55 ± 9 47 ± 5

III T2(ms) A(%) 115 ± 5 24 ± 8 121 ± 20 45 ± 9 105 ± 16 34 ± 6

143 Multicomponent T2 relaxation of human median nerve in vivo G. Gambarota1, T. W. J. Scheenen1, B. E. Cairns2, D. W. J. Klomp1, A. Heerschap1; 1Department of Radiology, UMCN, Nijmegen, NETHERLANDS, 2Faculty of Pharmaceutical Sciences, The University of British Columbia, Vancouver, BC, CANADA. Introduction: Measurements of transverse magnetization relaxation decays offer a method to study biophysical properties of nerves and could offer a means to investigate neurodegenerative diseases [1]. In vitro and in vivo animal studies of the proton transverse relaxation of peripheral nerves revealed multiple T2 components that are thought to result from spatial compartmentation of water into myelin, interaxonal and intraxonal spaces [2-4]. To date, the T2 relaxation decay of peripheral nerves has not been investigated in humans. In the present study, we measured the T2 decay of the median nerve in human voluteers. Methods: Magnetic Resonance Imaging (MRI) experiments were performed on a clinical 1.5T Siemens Sonata scanner. The body coil of the scanner was used as a transmitter and a surface coil (7x4cm) placed over the wrist in three volunteers acted as a receiver. Axial gradient echo multislice images were acquired to identify the median nerve. Multi echo imaging (32 echoes, TR/TE=3000/8.2ms, voxel-size=0.78x0.78x3mm) was then performed on a slice through the wrist. The images were analyzed on a PC station with Matlab. The non negative least square (NNLS) and Levenberg Marquardt (LM) non linear least squares algorithms were used to analyze the multicomponent T2 tissue relaxation data. Results: Figure 1 illustrates the location of the median nerve in an axial image of the wrist. The image-based T2 relaxation decay of the median nerve was multiexponential. The NNLS analysis of the decay data revealed the presence of two relaxation components at approximately 20 and 100ms (Figure 2, top). The LM biexponential fit of the same T2 decay data (Figure 2, bottom) yielded 22 and 110ms for the T2 values and 75% and 25% for the normalized amplitudes of the two components, respectively. The average and comparison with previous studies is shown in the table. Discussion: In this study, in vivo measurements of T2 relaxation decay in median nerve revealed that there were two components for the image derived T2 decay curves. This finding indicates the

144 Fat pad displacement in the feet of diabetic patients with neuropathy and claw toe deformity S. A. Bus1, M. Maas2, R. P. J. Michels1, M. Levi1; 1Internal Medicine, Academic Medical Center, University of Amsterdam, Amsterdam, NETHERLANDS, 2Radiology, Academic Medical Center, University of Amsterdam, Amsterdam, NETHERLANDS. Introduction: Plantar ulceration is one of the most serious complications of the foot in diabetic patients with peripheral neuropathy. Elevated dynamic plantar pressure is a major risk factor for ulceration. Claw toe deformity has been suggested to increase

S76

Functional MRI

plantar pressures by displacing the protective sub-metatarsal head (MTH) fat pad distally (Boulton, 1988). However, objective quantitative evidence for such an association does not exist. Therefore, the purpose of this study was to use MRI to assess the relationship between claw toe deformity and sub-MTH fat pad status in diabetic neuropathic patients. Subjects and Methods: Thirteen diabetic patients (8 men, mean age 56.3 yrs.) with sensory neuropathy and claw toe deformity and 13 age and gender matched diabetic neuropathic controls without deformity participated. A 1.5 Tesla magnetic resonance imager (Vision, Siemens, Germany) was used to acquire high-resolution (512x512 pixels) T1-weighted spin-echo sagittal plane images of the foot. In these images, fat tissue thickness was measured from regions located plantar to the MTH and the proximal phalanx. The ratio of these two thickness measures was used as indication for fat tissue displacement. Results: The mean (sd) sub-MTH fat pad thickness was 2.7 (1.5) mm in the patients with deformity and 6.3 (1.8) mm in the controls (P < 0.005). In six subjects with deformity, fat tissue was completely absent from this region. The sub-phalangeal tissue thickness in the group with toe deformity was 10.9 (2.4) mm compared with 8.9 (1.4) mm in the controls (P < 0.05). As a result, the thickness ratio was significantly smaller in the subjects with deformity (0.26 (0.16) vs. 0.70 (0.16), P < 0.005) indicating distal fat pad displacement in the deformed feet. Discussion/Conclusion: These results suggest a significant compromise in shock-absorbing capacity underneath the MTHs in feet with claw toe deformity. This is expected to result in elevated plantar pressures and thus a higher risk for plantar ulceration in diabetic patients who have lost protective sensation. MRI may be used to define areas at risk for development of ulceration. References: Boulton, AJM [1988] Med. Clin. North Am. 72:1513-1530.

145 In vivo MR elastography of skeletal muscle K. Uffmann, S. Maderwald, S. Mateiescu, W. Ajaj, C. J. Galbán, M. E. Ladd; Department of Diagnostic and Interventional Radiology, University Hospital Essen, Essen, GERMANY. Introduction: MR Elastography (MRE) is a non-invasive method for measuring the elasticity of human tissue. The feasibility of performing these measurements in skeletal muscle has already been established [1]. Currently there is interest in determining if muscle tissue changes due to pathological processes yield an effect that is detectable by MRE. The goal of this work was to establish an elasticity database for various muscle groups in healthy subjects. Methods: Mechanical waves were induced into the region of interest with a piezoelectric oscillator [2], and a phase-contrast MRI sequence synchronized to the oscillation was used to visualize the mechanical waves in the tissue. The shear modulus was determined from the wavelength inside the tissue using multiple phase offsets and motion encoding in all 3 spatial directions. The oscillator was coupled to a tendon attached to the muscle of interest (Fig. 1). Frequencies of 142 Hz and 100 Hz were used for the upper and lower extremity muscles, respectively. The amplitudes were in the range of 600 to 700 µm. Coronal slices along the muscle were acquired in a total scan time of approximately 30 minutes per muscle group. Measurements were performed on the biceps, soleus and gastrocnemius of 12 and in the flexor digitorum profundus of 4 healthy young volunteers using a 1.5T Siemens Sonata scanner. Results: All volunteers tolerated the examination. An average elas-

ticity value, represented by the shear modulus, of 62 (± 29) kPa, 47 (± 7) kPa, 38 (± 16) kPa, 52 (± 50) kPa was determined for the biceps, flexor digitorum profundus, soleus, and gastrocnemius, respectively. Discussion: MRE was used in this study to provide baseline measurements of muscle elasticity in various muscle groups in healthy human subjects. To determine the efficacy of MRE as a viable diagnostic/prognostic method, diseased muscle tissue must be examined to compare its elasticity with the presently acquired data. Additionally, further measurements in a group of volunteers should be performed to prove the reproducibility of the MRE examinations. To better evaluate muscle elasticity, other MRI measurements, such as muscle volume, diffusion and T2 relaxation, could be of value. References: 1. Dresner et al., JMRI 13:296-276(2001) 2. Uffmann et al., Concepts of Magnetic Resonance (Magnetic Resonance Engineering) 15(4):239-254(2002)

Poster Walking Tour Functional MRI 2:30 pm - 3:30 pm

146 Separating the contributions to the intravascular signal change in Spin-Echo fMRI at 3 Tesla T. H. Jochimsen1, M. v. Mengershausen1, T. Mildner1, D. G. Norris2; 1Nmr, Max Planck Institute of Cognitive Neuroscience, Leipzig, GERMANY, 2Centre for Cognitive Neuroimaging, FC Donders, Nijmegen, NETHERLANDS. Introduction: Diffusion weighted fMRI (1,2,3) is a valuable tool to quantify the origin of the functional signal, i.e. contributions from the larger vessels, from the capillaries, or the extravascular space. The attenuation of the intravascular signal due to the application of gradient pulses can be explained by two mechanisms (4,5): The first component consists of spins that perform a random walk and is referred to as pseudo-diffusion. The second component can be modelled by a network of isotropically oriented pipes with laminar flow and a static velocity. To determine the contributions of these components to the functional contrast, the approach of flow rephasing and dephasing (6,7) was applied to fMRI.

Functional MRI Subjects and Methods: A functional study with 6 subjects and visual stimulation during the acquisition of spin-echo EPI images at TE=96.5ms and TR=1050ms was performed. Diffusion weighting was introduced into the sequence by two consecutive bipolar gradient pulses. A second variant of this single-shot acquisition scheme with flow-compensation was obtained by reversing the first bipolar pulse and measured subsequently. This scheme was used for 6 different b-values which were measured in an interleaved fashion together with a non-diffusion-weighted image. Results: A ROI that shows significant correlation (z>2.6) was selected from the non-diffusion-weighted measurement and used as mask to select the time courses of the diffusion-weighted interleaves within the same trial. The fractional signal change was then calculated by averaging the time courses for each b-value. The result in dependency of the b-value is shown in Figure 1. Discussion and Conclusion: Both curves show a large drop of the functional signal for small b-values. This can be assigned to spins showing a pseudo-random movement, e.g. turbulent and pulsatile flow. The decay of 0.5-0.6% for the flow-rephased signal can therefore be attributed to larger vessels where blood flow is fast. The constant difference of 0.3-0.4% between the two curves for higher b-values can be explained by a component with coherent flow. Because the difference remains constant over a wide range of b-values, this fraction may be assigned to smaller blood vessels with a smaller velocity. These data suggest, that a b-value of 50 s/mm² without flow-compensation is sufficient to suppress nearly all of the intravascular contribution to the functional signal. References: 1. Boxerman JL, Bandettini PA, Kwong KK, Baker JR, Davis TL, Rosen BR, Weisskopf RM [1995] Magn. Reson. Med. 34:4-10 2. Song AW, Wong EC, Tan SG, Hyde JS [1996] Magn. Reson. Med. 35:155-158 3. Norris DG, Zysset S, Mildner T, Wiggins CJ [2002] NeuroImage 15:719-726 4. Le Bihan D, Breton E, Lallemand D, Aubin ML, Vignaud AJ, Laval-Jeantet M [1988] Radiology 168:497 5. Henkelman RM, Neil JJ, Xiang Q [1994] Magn. Reson. Med. 32:464-469 6. Ahn CB, Lee SY, Nalcioglu O, Cho ZH [1987] Med. Phys. 14(1):43-48 7. Fujita N, Harada K, Sakurai K, Akai Y, Kozuko T [1992] Magn. Reson. Med. 24 :109-122

S77

147 Towards Spin-Echo BOLD fMRI using SENSE at 3 Tesla S. Boujraf1, P. Summers1, K. P. Pruessmann2, S. Kollias1; 1Institut für Neuroradiologie, UniversitätsSpital Zürich, Zürich, SWITZERLAND, 2Institute for Biomedical Engineering, ETH, Zürich, SWITZERLAND. Introduction: The choice of BOLD-fMRI technique may impact upon the accuracy of functional localization accuracy by affecting image quality and the degree of BOLD contrast achieved (1). By improving samling efficiency, parallel acquisition techniques such as SENSE have been used to shorten readout trains in single-shot gradient echo EPI, thereby reducing artifacts and improving spatial resolution (2). SENSE together with single-shot spin-echo (SS-SE) EPI imaging may further reduce off-resonance artifacts. The goal of this work was to investigate the BOLD response of a SENSE adapted SE-EPI on 3 Tesla scanner. Materials and Methods: fMRI studies of 11 healthy right hand dominant volunteers were carried out in 3 Tesla scanner (Gyroscan Intera, Philips Medical Systems) using the body coil for rf-excitation and a 6-element head coil as receiver. fMRI of the whole brain was performed using a SS-SE EPI sequence (4 mm thick, 0.4 mm slice gap, FOV of 240 mm, 128x128 recon matrix, TR 2400 ms, flip angle 90°) for each of the following TEs: 35, 40, 45, 49 ms. A SENSE factor of 2.75 was used reduce the number of phase encoding steps. The subjects performed a self-paced (~1.5Hz), simple motor task with the dominant hand for 30 seconds alternating with rest in a block design for 3 minutes (72 time points). All data underwent identical post-processing (3). Only pixels with a statistically significant correlation (P<0.001) were considered as activated areas. Average and maximum BOLD contrasts were determined for the cluster of activation in the M1 motor area in each subject. Results: Three primary remarks can be made. First, quadrature ghost artifacts, which are typically seen in single-shot EPI, were not visible. Secondly, the distribution of activation in the motor, pre-motor and sensory region (Figure 1) was consistent with previously reported studies for this motor task. Lastly, the level of BOLD contrast increased with increasing echo time (Table 1) consistent with the expected time dependence of the BOLD phenomenon. Discussion and Conclusion: SE-EPI using SENSE at high magnetic fields can be used to investigate functional brain activity. As noted by previous authors in the context of gradient echo EPI, susceptibility artifacts can be reduced by use of SENSE. At high field strengths, such measures are crucial. SE-EPI with SENSE may offer further benefits in this regard as well as being more specific to oxygenation changes in the microvasculature. References: 1. MRM 1995;34:555-566 2. MRM 2001;46:638-651 3. MRM 1993;30:161-173 Table 1. Average and maximal change of BOLD signal with ncreasing echo-time

Echo time TE in ms Average Percentage BOLD signal change Maximal Percentage BOLD signal change

35

40

45

49

0.93

1.27

1.78

1.47

3.3

3.5

3.8

4.3

S78

Functional MRI

Summary of Scanner Performance

Scanner A B C D

%S 0.39 0.50 0.64 1.06

GSR 0.04 0.03 0.23 0.40

Fn 0.0013 0.0015 0.0016 0.0025

148 Inter and Intra-System Quality Assurance of fMRI G. P. Liney, R. Garcia-Alvarez, L. W. Turnbull; Centre for MRI, Hull Royal Infirmary, Hull, UNITED KINGDOM. Introduction: Functional MRI (fMRI) is being increasingly used for surgical and radiotherapy planning [1], thereby increasing the demand for a robust and reliable technique. This work describes a quality control investigation of fMRI sequences on different MRI systems. Methods: Four 1.5 Tesla scanners (Clinical scanner A, Research scanner pre- and post-upgrade (B & C), and a mobile D) were assessed over a period of several months. Images were acquired in a suitable phantom using a standard fMRI protocol, consisting of a single-shot gradient-echo EPI sequence (TE/TR/flip = 40 ms/3 s/90°). A total of 70 phases were acquired at each of six axial slices (5/1.5 mm). Scans were repeated to access short term stability. Isocentre images were assessed in terms of distortion, maximum ghost-to-signal ratio (GSR) and maximum percentage signal change (%S). Signal fluctuation, Fn [2], was calculated for various pixels sizes (n = 3, 5, 10). The data was also evaluated to investigate any significant variation with time. Results: Figure 1 illustrates images acquired with scanners D and B exhibiting different levels of distortion and ghosting. Results for each scanner are summarised in the table which gives the largest values recorded during the course of the study. Scanners A and B consistantly demonstrated long term stability (%S < ±0.5%) over several months. Following upgrade, scanner C demonstrated increased signal fluctuation and poor ghosting. Scanner D produced the worst results overall. Additionally, both B & D demonstrated a significant signal drift over the time course (P <0.001). Figure 2 illustrates signal change over the time course for scanner A (triangles) and B (squares); the former demonstrating the least fluctuation and no significant trend with time, while the latter clearly shows a significant signal drift.

Conclusion: This work demonstrates the short and long term variability in performance of clinical MR systems and the need for quality assurance in fMRI studies. Our results indicate that not all of the scanners achieved a stability criterion of ± 0.5 % at each of the examinations. The results stress the importance of adopting this procedure both as a routine quality control check and as part of the acceptance testing of any new scanner installation or upgrade. References: 1. R Garcia-Alvarez, GP Liney et al Journal of Radiotherapy in Practice 2003; 3(2):63-69. 2. RM Weisskoff Magn. Reson. Med. 1996; 36: 643-645.

149 Application of fMRI in estimation of neural processing of the affective visual stimuli A. Urbanik1, B. Sobiecka1, J. Kozub1, L. Podsiadlo1, M. Kuniecki2; 1Department of Radiology, Jagiellonian University, Collegium Medicum, Kraków, POLAND, 2Department of Psychophysiology, Jagiellonian University, Collegium Medicum, Kraków, POLAND. Purpose: The central mechanisms underlying emotional appraisal of incoming stimuli has been extensively studied both in animal and human subjects. Despite the fact that understanding of this issue is increasing, some specific questions remain to be answered. In the following study, using fMRI technique, we examined how the brain processes emotionally loaded (positively and negatively) visual information. Materials and Methods: Seventeen volunteers were examined in the Signa 1.5T MR system (BOLD). Each experimental session

Functional MRI consisted of five activation periods, three of which constituted baseline and two experimental condition. In experimental condition subjects were shown slides selected on the basis of high score on Arousal and Valence scale from standardized IAPS picture set, while in the baseline condition custom prepared color checkerboards were presented. There were two runs, in the first run subjects saw only negatively valenced pictures, during the second run only positive pictures were shown. Results: For the positive stimuli we detected significant activation in the bilateral occipito-temporal cortex, as well as left anterior insular cortex. For the negative stimuli, in addition to the previous structures, we also observed activation of the right amygdala. Conclusion: Activation of bilateral occipito-temporal cortex proves enhancement of visual processing of negative slides as compared to neutral checkerboards. This might be attributed to topdown processes. Such hypothesis is further supported by apparent activation of amygdala which maintain connections with visual cortex. Amygdala is a structure commonly described as being involved in processing of negative stimuli. Activation of insular cortex is probably related to autonomic arousal accompanying watching emotional content.

150 FMRI of the human sensorimotor cortex before and after subsensory whole-hand afferent electrical stimulation S. Golaszewski1, C. M. Siedentopf2, F. Koppelstaetter3, G. M. Guendisch3, M. Verius3, F. M. Mottaghy4, M. R. Dimitrijevic5, F. Gerstenbrand6, S. R. Felber7; 1Neurology, University of Graz, Graz, AUSTRIA, 2Radiology II, University of Innsbruck, Innsbruck, AUSTRIA, 3Radiology II, University of Innsbruck, A-6020 Innsbruck, AUSTRIA, 4Nuclear Medicine, Research Center Juelich, D-Juelich, GERMANY, 5Division of Restorative Neurology and Human Neurobiology, Baylor College of Medicine, Houston, TX, 6Ludwig Boltzmann Institute for Restorative Neurology and Neuromodulation, Otto Wagner Hospital, Vienna, AUSTRIA, 7Neuroradiology, University of Innsbruck, A-6020 Innsbruck, AUSTRIA. Introduction: Stimulation of propriozeptive pathways using whole-hand electrical stimulation with a mesh-glove has been shown to improve motor performances of stroke patients with chronic neurological deficits. The aim of the study was to elaborate, whether changes in the motorcortex activation pattern can be demonstrated after electrical stimulation of the hand in volunteers. Materials and Methods: All experiments were performed on a 1.5Tesla MR-scanner in 10 healthy subjects. The motor-paradigm was self-paced finger-to-thumb-tapping of the left hand. Firstly, a baseline fMRI-examination and secondly subthreshold electrical stimulation with 0.9mA was applied for 20 minutes outside the magnet to the left hand using a mesh-glove. Thirdly, an identical fMRI run to the baseline and the second run 12 hours post stimulation was performed. Post processing was done with SPM99. Results: Group-analysis of fMRI-data showed: 1. Baseline fMRIexaminations revealed brain activation of the primary and secondary sensorimotor cortex as previously described. 2. After electrical stimulation of the left hand, there was an increase of activated pixels in these areas. 3. In addition, there was activation of regions not visible on the baseline studies. These involved the ipsilateral inferior parietal lobule, the pre- and postcentral gyrus and the superior parietal lobule. 4. These changes disappeared twelve hours post stimulation.

S79

Conclusions: fMRI reflects an increased BOLD-response due to an increase of local-field-potentials within the sensorimotor cortex as a consequence of electrical stimulation. Thus, local-field-potentials can be successfully influenced by subsensory stimulation of afferent pathways. This holds promise for the application of fMRI in the planning of neurorehabilitation strategies.

151 Ipsilateral cortical activations induced by laser acupuncture in humans C. M. Siedentopf1, S. M. Golaszewski2, F. Koppelstaetter1, G. M. Guendisch1, P. Rhomberg1, F. M. Mottaghy3, A. Schlager4, S. R. Felber1; 1Radiology II, University of Innsbruck, Innsbruck, AUSTRIA, 2Neurology, University of Graz, Graz, AUSTRIA, 3Department of Nuclear Medicine (KME), Research Center Juelich, Juelich, GERMANY, 4Department of Anaesthesia, University of Innsbruck, Innsbruck, AUSTRIA. Purpose: Recent functional brain imaging studies could demonstrate the effect of needle acupuncture on activities of specific brain areas. There were first notes for an effect of lateralisation of cortical activations on the ipsilateral cortex hemisphere on the bodyside to which acupuncture was performed. Thus we wanted to examine this effect. We chose acupoint GB43 (Xiaxi) on the foot and compared the results of the laser-acupuncture of the right acupoint GB43 with the results of the left acupoint GB43. The traditional indications to use GB43 are deafness, tinnitus, dizziness, eye diseases and building up willpower. Methods: We analysed 2 groups containing 11 healthy male volunteers (age: 20–38) each, one group for the right and one for the left acupoint. We chose acupoint GB 43 which is located between the fourth and fifth toe. As a control condition the head of the laser was mounted on the same acupoint resulting in the same tactile stimulus as in the verum condition, but without application of active laser stimulation. Volunteers were not able to differentiate between verum and placebo acupuncture. All experiments were performed on a 1,5T MR-Scanner. For fMRI, we employed T2*weighted single shot EPI-sequences (TR/TE/α=0,96ms/66ms/90o, matrix=128x128). Post-processing was performed on a workstation with SPM99. Results: The group results of the left acupoint GB43 showed significant activations in the substancia nigra, nucleus rubber, brainstem and the gyrus cinguli (BA24) on the left side. No activation on the right side could be detected. The group results of the right acupoint GB43 showed significant activations in the right brainstem, the right lobus parietalis inferior (BA40) an the left and right gyrus frontalis (BA9). The results of the placebo groups of the left and right acupoint GB43 revealed no significant activations. Conclusion: This study revealed primarily activations in the ipsilateral cerebral hemisphere with respect to the acupoint. The fact that the placebo acupuncture showed no effect underlines the specificity of this observation which is in accordance with the findings of our former study [1]. Hence we could show once more that acupuncture activates predominantly ipsilateral brain areas. One might hypothesize that ipsilateral activation are an indication that acupuncture is mediated by meridians, since meridians do not cross to the other side like the afferent nerves. References: 1. Siedentopf CM et al. fMRI detects activation of the visual cortex during laser acupuncture of the foot (BL 67) in humans; Neuroscience Letters, 2002 Jul 12;327(1):53-6.

S80

Functional MRI

152

153

Functional magnetic resonance imaging (fMRI) in the treatment planning of therapy-resistent neurological painsyndromes with chronic motor cortex stimulation F. Koppelstaetter1, S. Golaszweski2, P. Rhomberg1, C. M. Siedentopf1, G. M. Guendisch1, W. Recheis1, T. Fiegele3, F. M. Mottaghy4, S. R. Felber1; 1Radiology II, University of Innsbruck, Innsbruck, AUSTRIA, 2Neurology, University of Graz, Graz, AUSTRIA, 3Institute for Ion Physics, University of Innsbruck, Innsbruck, AUSTRIA, 4Department of Nuclear

Basal ganglia activity during internal generation of action C. Windischberger1, R. Cunnington2, E. Moser1,3; 1Institute of Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Howard Florey Institute, University of Melbourne, Melbourne, AUSTRALIA, 3Department of Radiodiagnostics, University of Vienna, Vienna, AUSTRIA.

Medicine (KME), Research Center Juelich, Juelich, GERMANY. Purpose: Treatment of therapy-resistant neurological painsyndromes is difficult. Neurosurgery offers different methods for treatment. The problem is the correct electrode placement. The electrode has to cover the pre- and postcentral cortex respecting the somatotopy of the homunculus and the patients pain distribution. We used fMRI to plan the position of the implantation of the electrodes and intraoperative transdural electrical stimulation for functional reevaluation. Our objective was to demonstrate the combination of both methods for treatment of therapy-resistant neurological pain-syndromes with chronic cortex stimulation. Methods and Materials: We treated 4 patients with a chronic pain-syndrome following traumatic low cervical cross-section (n=1), cervical roots avulsion (n=1), denervation of the lumbar plexus (n=1) and left sided thalamic infarction (n=1) with chronic epidural motor cortex stimulation. For fMRI we used standard motor and somatosensory paradigms on a 1.5 Tesla MRI-Scanner (Magnetom VISION, Siemens, Germany) with an echo-planar capable gradient system (rise time 300µsec, 25mT/ms) and a circular polarized head coil (FoV=240mm). We employed T2* EPI-sequences (TR/TE/α=0,96ms/66ms/90o, matrix=64x64, acquisition time: 2 sec, voxel dimension of 4x4x4mm). We acquired 24 slices parallel to the bicommissural plane. (TR=4 sec). Post-processing was done with SPM99. A statistical parametric activation map for each single subject was performed using a corrected p value <0.001 and a cluster size of 4. For IOM prior to the neurosurgical electrode implantation the exposed area was stimulated by utilizing a bipolar stimulation electrode using a Nicolet Viking IV (Nicolet, USA) device (50 Hz square wave impulse, 0.2 msec duration, 5-15 mA intensity). The patients were operated under oral intubated anesthesia without relaxation. Results: In fMRI sensory and motor cortical brain areas could be well differentiated within both hemispheres in all patients. Postoperatively two patients had a complete loss of pain. The patient with the thalamic infarction had no more facial pain. One patient had no positive pain-reductive result despite of correct electrode position. Conclusion: Our results demonstrate that fMRI can be used for preoperative planning of neurosurgical chronic motor cortex stimulation. It delivers additional information for the intricate placement of the electrodes. It has to be further evaluated whether specific activation patterns seen in a fMRI study could be predicitive for therapy outcome.

Introduction: The internal representation of an action that is about to be performed is extremely important for fast and duly execution. Recent studies have suggested that when someone observes a person performing an action, the observer will form an internal representation of that action as if the observer would be actually performing this action [1]. In this study, we have examined differences in brain activity for observed vs. internally generated movements during presentation, preparation and execution. Subjects and Methods: Fourteen young, male subjects have been studied on a 3 Tesla Bruker Medspec scanner (Bruker Biospin, Germany) using gradient-recalled EPI (17 slices, TE/TR=35/1000). Prior to the measurement subjects extensively trained a set of six movements taken from the American sign language alphabet. During the experiment subjects were either shown a 2sec movie of a novel sign language letter and were asked to closely observe this movement and repeat it after 5sec, or subjects were presented with a heavily filtered movie were no action was recognizable and were requested to perform one of the six trained gestures instead. Analysis was done in SPM applying a finite impulse response (FIR) approach to examine the temporal evolvement of brain activity. Results: The figure shows four slices of the random effects group analysis results for the observation/self-generating part of the experiment over all fourteen subjects (p<0.001, uncorrected) overlaid to the mean EPI data set: (a) Voxels with greater activity during internal movement generation compared to action observation, (b) vice versa. Self-generated actions show increased activity in the head of the caudate nucleus (arrows), as well as pre-frontal areas. Observing a movement causes neural activity in higher visual areas and parietal regions. These differences are still present during the first 2sec after stimulus presentation, but diminish later on. Discussion: Our results show that internal generation of movements requires significantly more basal ganglia activity than observing a movement. The primary activation foci within the basal ganglia have been found in the head of the caudate nucleus which ,as part of the striatum, is a major recipient of cortical input. These results highlight the important role of the basal ganglia in the internal generation of action. References: 1. Jeannerod M, Quarterly Journal of Experimental Psychology, 1999, 5A: 1-29. Acknowledgements: This work has been supported by the Austrian National Bank (ÖNB-JF9305) and the Hochschuljubiläumsstiftung of the City of Vienna (1472/2002).

Functional MRI 154 The role of the suprapontine brain structures in voluntary control of voiding - A high field fMRI study H. Zhang1, A. Reitz1, A. Curt2, S. Kollias3, B. Schurch1; 1Institute for Rehabilitation and Research, Department of Neurourology, University Hospital Balgrist, Zurich, SWITZERLAND, 2Institute for Rehabilitation and Research, Department of Neurology, University Hospital Balgrist, Zurich, SWITZERLAND, 3Institute for Neuroradiology, Department of Neuroradiology, University Hospital Zurich, Zurich, SWITZERLAND. Introduction: Despite the long time speculation from clinical observation that there is inhibitory influence of frontal lobe on micuturition, and the major inhibitory brain ares are located in "medial prefrontal cortex" and "basal ganglia", yet not too much so far has been done to clarify the suprapontine structures involved in voluntary control of visceral function. The aim of this study is to explore the brain structures involved in the voluntary control of the voiding, in other term, to answer the question: "Who is keeping the detrusor from contraction and sphincter from relaxation until the micturition can be started at socially appropriate time?" Method: Philips 3 T Gyroscan (Located in Institute of Biomedical Engineering. ETHZ, http://www.mr.ethz.ch/3t/) was used in this study. All the images were transferred to a Linux server via DICOM node for post-processing using SPM99 (Wellcome Department of Cognitive Neurology, London) Since May 2002 to March. 2003, 12 right-handed male students (mean age of 23.8 yrs, SD=0.65) were recruited from UniZurich medical school. The study details and paradigm as shown in table1 (15 seconds off-on block design, 4 mins) were explained to all participants and pelvic floor contraction was tested and practiced with a biofeedback device if needed (SEDIA Pelvitrain®, Sedia AG, Switzerland). All the subjects read and signed informed consent prior to the scan. Data post-processing: 3 subjects' data sets were excluded from the 11 volunteers for the corrupted images due to wrong acquisition procedure or other unknown reasons. 8 subjects' data motioncorrected and normalized before statistical analysis. After reviewing each individual's T-map, a group conjunction analysis was performed for the group full-bladder pelvic floor contraction and group empty-bladder pelvic floor contraction. Furthermore a double subtraction [(A - B)-(C - D)] paired T test was performed between these two groups to observe the net inhibitory effect induced by pelvic muscles contraction. Results: Our results demonstrate that the supplemental motor area (SMA, BA6) and bilateral putamen in basal ganglia and cerebellum (figure 1-2) are the major activation areas for the inhibitory bladder control center as seen from our healthy volunteer experiment scenario. Discussion: The present finding confirmed "medial prefrontal cortex" and "basal ganglia" are indeed involved in inhibitory bladder control,which has long been long speculated by clinical reports derived from stroke patient urinary syndrome and brain images data, and proves the potential value of high field fMRI in clinical application in our spinal cord injury patient in the future.

S81

Table 1 (Block design off-on paradigm) Sensory-motor events

Sensory-motor events

during pelvic floor

during pelvic floor

contraction with full

contraction with empty

bladder, 4 mins in total

bladder, 4 mins in total

Pelvic floor contraction

Pelvic floor muscle

Pelvic floor muscle

(on) 15 seconds

contraction, bladder

contraction

sensation, desire to void,

[condition C]

inhibition of voiding [condition A] Pelvic floor ontraction

bladder sensation

NULL

rest (off) 15 seconds

[condition B]

[condition D]

Net effect compared by

Pelvic floor contraction,

Pelvic floor muscle

SPM between the two

inhibition of voiding

contraction

groups, double contrasts:

[A-B]

[C-D]

[(A-B)-(C-D)],observating inhibition of voiding

155 In vivo and ex vivo characterization of an USPIO (Sinerem) and applications to fMRI in rat A. Adami1, P. Fantazzini2, C. Garavaglia2, E. Nicolato1, J. Tommasi3, L. Calderan1, P. Marzola1; 1Morphological and Biomedical Sciences, University of Verona, Verona, ITALY, 2Physics, University of Bologna, Bologna, ITALY, 3Dipartimento Provinciale di Verona, ARPAV, Verona, ITALY. Introduction: Ultrasmall Superparamagnetic Particles (USPIO),

S82

RF-System

originally developed for liver imaging, are now used in a number of applications including fMRI (1). Here we show the use of a commercially available USPIO (Sinerem, Guerbet France) for fMRI in rat. The contrast agent relaxivities in plasma and blood-half times have been characterized ex vivo and in vivo. Subjects and Methods: Ex vivo experiments: Sinerem was administered i.v. to Wistar rats (n=7) at a dosage of 6mgFe/kg. Blood samples were withdrawn at different time points after injection (from 10 to 360 min) and plasma fractions were obtained by centrifugation. T1 and T2 relaxation times were measured by IR and CPMG pulse sequences, respectively, using a Minispec P20 (Bruker, Karlsrhue, Germany) operating at 20 MHz and 37.0 ± 0.1°C. The iron concentration was determined by Atomic Absorption Spectroscopy. fMRI experiments were performed using SD rats (n=4). The stimulation consisted of 3ms rectangular pulses, 10-15V amplitude and 3Hz frequency applied to the left forelimb. The fMRI activation paradigm consisted of an Off-On-Off sequence. Images were acquired using a Bruker Biospec (4.7T) and a GRE sequence. Results: The r2 relaxivity of Sinerem measured ex vivo in rat plasma decreases with the time elapsed between injection and withdrawing: r2=83.7 (mmol/L)-1s-1 and r2=35.3 (mmol/L)-1s-1 10 and 360 min after injection, respectively. During the same time interval r1 also decreases but at a lesser extent (38% against 58%): from 26.5 (mmol/L)-1s-1 to 16.4 (mmol/L)-1s-1. This is probably due to interactions between the contrast agent and blood components or to degradation of the contrast agent itself. Interestingly the dependence of the tranversal plasma relaxation rate (R2) on iron concentration (C) is better described by a non linear (R2= 45.1*C1.64) than by a linear relationship; similar results have been recently reported for R2* in a different USPIO (2). In vivo and ex vivo blood-half times were in agreement with the literature. Functional MRI experiments showed a (9±1)% decrease in the signal intensity of the activated pixels, significantly higher than the effect observed by BOLD. Conclusions: This study shows the usefulness of a commercially available USPIO in fMRI. Since the relaxivities of this contrast agent in plasma are not constant, this study also indicates that extracting USPIO concentration values from Signal Intensity (or relaxation time) measurements is potentially incorrect. References: 1. Dijkhuizen RM et al, PNAS 98:12766-71 (2001) 2. Bjornerud A, et al, MRM 47:461-471(2002).

Poster Walking Tour RF-System 2:30 pm - 3:30 pm

156 Considerations for appropriate volumetric array designs with many channels for high-performance PPA M. F. Mueller, M. A. Griswold, A. Webb, A. Haase, P. M. Jakob; Physikalisches Institut, Experimentelle Physik 5 (Biophysik), Würzburg, GERMANY. Introduction: Recently large receiver subsystems have been developed [1]. The goal of this abstract is to determine which coil design could be used for volume-arrays with 32-64 receiving elements. We simulated, constructed and analyzed various designs which could potentially be used for many-channel head-arrays.

It has been shown that 1D-acceleration in the transverse direction is limited regardless of the coil arrangement. Therefore we have focused on designs which can encode in more than one direction for multi-slice and 3D applications, considering birdcage-like volume array coils, including spiral surface-arrays (Figure1). Methods: Several arrays of the types shown in Figure1 were simulated. Geometry-factor maps were calculated for several encoding directions using a SENSE reconstruction. These maps could be used as criteria for judging the encoding performance of the different arrays. To determine the overall SNR performance of each array, the intrinsic SNR of the array is needed. To determine this, single coil elements of various arrays were constructed on a cylindric carrier. To evaluate the performance of the coils, the unloaded and loaded Q-factors and the B-field strength in the middle of the cylinder was measured for each coil. Results: G-factor maps for axial-encoding for the two different coil arrangements (Figure1) are shown in Figure2. The spiral has approximately 40% worse g-factor performance than the 2-ring birdcage, although axial-encoding is not possible for usual singlering birdcages. The spiral showed an advantage in terms of Q and B1-field compared to the birdcage. An additional gain in intrinsic SNR could come from adding an extra spiral with a twist of -0.5π, since the noise correlation with the +0.5π spiral should be small [2]. The benefit in B1-field in the center compared to the birdcage could be of 5-6 dBs at a corresponding level of coil losses, which would more than make up for any g-factor losses in the spiral array. 8 element 1-ring birdcage 16 element 1-ring birdcage 32 element 1-ring birdcage 8 element +0.5π spiral

B1 -27 dB -32 dB -37 dB -29 dB

Qu 320 268 256 375

Ql Qu / Ql 120 2.67 120 2.23 168 1.52 150 2.50

Conclusions: Several coil array geometries were examined for their potential use in arrays with large numbers of coil array elements. While the spiral arrays have a higher g-factor, we predict that the gains in intrinsic SNR should more than make up for these losses. Another aspect we didn’t examine in detail is our ability to build these arrays. The 32-element birdcage is already more copper area than open flux area, so we believe the spiral arrays should scale upwards better than the straightforward birdcage arrays. References: 1. Brown, ISMRM2002, p.863 2. Duensing, ISMRM2002, p.771

RF-System

157 Multi-channel resonators for experiments at 17.6T V. C. Behr, M. Oechsner, D. Gareis, I. Wieland, T. Neuberger, A. Haase, C. Faber; Department of Physics, EP5 (Biophysics), Würzburg, GERMANY. Introduction: High field spectrometers operating at 17.6T for small animal imaging have recently become available. Microscopic imaging as well as spectroscopic experiments benefit from improved SNR and increased chemical shifts yielded by these field strengths. At frequencies up to 750MHz, the larmor frequency of 1H at 17.6T, established hardware designs have to be re-evaluated. In this abstract we present the results of our preliminary studies with double-resonant and quadrature resonators for 1H, 19F and 23Na MR imaging and spectroscopy. Setup: All experiments are performed on a Bruker Avance wide bore magnet at 17.6T. This system can be equipped with either a 1T/m gradient unit with an inner diameter of 40mm or one with 200mT/m and 56.5mm inner diameter. Furthermore a choice of transmitters is available allowing 1H and X-nuclei studies. Hardware: In figure 1 two resonators built for multi-channel applications are shown. Both resonators are designed for the 200mT/m in-vivo gradient system and have an inner diameter of 38mm. The quadrature birdcage is an eight-legged high-pass birdcage with a height of 42mm. During its assembly even minor deviations of the capacitors’ values lead to major imbalances in the current distribution and, as a result, in the B1 field, since at a frequency of 750MHz capacitances were on the order of only a few pF. The double-tuned saddle coil pair is optimized for 23Na imaging with saddles of approximately (40mm)2 while the 1H saddles are about 30mm by 40mm and primarily intended for positioning of the sample and shimming. For this resonator the coupling between the overlapping saddles proved to be the main issue to overcome. In figure 2 an optimized layout for a double-tuned fourlegged birdcage resonator is shown. For the second channel the feeds as well as tuning and matching capacitors (CT and CM) are located 90° rotated with respect to the first (not shown). The common ground-ring is designed for stabilizing and decoupling of the two channels. Experiments and Conclusions: In figures 3a and 3b images recorded simultaneously with the two quadrature modes of the birdcage resonator are shown. As can be seen both channels yield good signal over the entire sample. Figure 4 shows a well resolved sodium image acquired with the saddle coil after positioning and shimming was done with the proton resonance (images not shown). Both resonators offer sufficient decoupling of their modes and B1 homogeneity to acquire highest quality images.

S83

S84

RF-System

158 An investigation of coil isolation techniques with high levels of mutual coupling D. J. Gilderdale; Robert Steiner MR Unit, Imperial College, London, UNITED KINGDOM. Introduction: Array coil design would be simplified if the effects of mutual coupling could be ignored. Roemer et al [1] demonstrated two methods which reduce coupling effects, but highlighted their limitations. Further improvements have been achieved by increasing the output impedance of the coil [2] and by employing lower input impedance preamplifiers [3,4]. We have investigated the performance of these techniques, employed separately and in combination, in the context of high mutual coupling. Methods: A loaded 12cm x 8cm 0.5T surface coil (A), matched to 50ohms with a 5ohm input impedance noise-matched preamplifier was driven by a 1cm diameter magnetic probe fixed at its centre. A second, identical coil/preamplifier assembly (B) was positioned coaxially with (A) and the SNR performance of (A) was measured, as a function of coil spacing, using a spectrum analyser. This experiment was repeated for coils matched to 850ohm output impedance, again with noise-matched preamplifiers. Both experiments were then simulated using linear circuit modelling software. Finally, we modelled the effect of reducing the preamplifier input resistance. Results: Fig.1 shows the SNR produced by coil (A) as a function of its separation from coil (B). Performance using standard preamplifiers is demonstrated for both rf measurements and simulation. The additional reduction in coupled noise produced by ultra low impedance preamplifiers is simulated for both 50ohm and 850ohm systems. Fig.2 shows the coupling coefficient as a function of coil spacing, for reference purposes. Simulation and experiment showed close agreement. Conclusions: As expected, close coupling of coils, even when using standard mis-matched preamplifiers, can produce large SNR losses. These losses are predicted by consideration of coupled coil noise alone, taking no account of the degradation of preamplifier noise figure as coil source resistances fall. Matching coils to 850ohms output resistance yields an improvement similar to that achieved by a 50ohm system using an ultra low impedance preamplifier. Use of such a preamplifier, appropriately noise matched, with an 850ohm coil should yield a significant further isolation gain. References: 1. Roemer PB, et al [1990] Magn. Reson. Med. 16:192-225 2. Gilderdale DJ, Larkman DJ [2002] MAGMA 15.1:144 3. Ledden PJ, Inati S [2001] Proc. ISMRM 9:1117 4. King SB, et al [2002] Proc. ISMRM 10:330

159 Profile definition of C-FOCI Inversion-Pulses: Dependency on maximum amplitude multiplication factors A. Bongers, L. R. Schad; Abteilung Biophysik und med. Strahlenphysik, German Cancer Research Center, Heidelberg, GERMANY. Introduction: Several techniques in physiologic MR-Imaging are based on slice-selective inversion pulses. Especially quantification of tissue perfusion with arterial-spin-labelling sequences relies on sharp inversion profiles for preparation-pulses. Lack of profile sharpness e.g. leads to significant subtraction errors and hence to substantial errors in perfusion quantification. For some time adiabatic hyperbolic-secans-pulses (HS-pulses) have been in use to prepare magnetisation. These pulses are insensitive to B1-inhomogeneities and deliver sharper inversion profiles than non-adiabatic sinc-pulses. However, as maximum pulse-bandwindth is limited by transmitter-voltage, inversion sharpness of slice-selective HS-pulses is not perfect either. C-FOCI-pulses [1] use modulated gradients to reduce the maximum B1-amplitude necessary for inversion and promise sharper inversion profiles. However, this requires high gradient-strengths which can become a technical problem for thin inversion-slabs. In this study C-FOCI-pulses with different amplitude-multiplication-factors (FMAX) were implemented and resulting inversion profiles were compared. Methods: C-FOCI-pulses with FMAX of 2,4,6,8,10 and 16 were implemented on a 1.5T scanner (Siemens Symphony, Gmax=30mT/m). The modulation-function for RF-amplitude, RFfrequency and gradient-strength were derived from HS-pulses using the following equations.

with B1=RF-amplitude, δ_upper; ω=frequency-offset, G=gradientstrength of the FOCI-pulses; µβ/π=bandwidth, GHYPSEC=gradient-strength of the underlying HS-pulse. F(t): modulation-function for the FOCI-pulse. FMAX is the multiplication-factor for the HSgradient-borders to obtain the corresponding FOCI-gradient. An inversion-bandwidth of 2500Hz, (β=1300,µ=6) was used with a pulse-duration of 10.24ms, representing typical values for optimized HS-pulses on hole-body-scanners. Inversion profiles for all pulses were determined in a phantom study. The preparation-pulses

RF-System were incorporated into a FLASH-sequence (δ_upper; x=δ_upper; y=0.8mm, slice-thickness=3mm, TE=4.8ms, TR=2000ms) with readout orthogonal to the inversion slab (TI=200ms). Results: Fig. 1 shows the profiles of the transition regions from non-inverted to fully-inverted magnetisation for the HS-pulse and the implemented C-FOCI-pulses with FMAX=2,4,6,8,10 and 16. As shown, C-FOCI-pulses deliver much sharper inversion profiles than an HS-pulse with the same bandwidth. To quantify the dependency of inversion-profile-sharpness on amplitude-multiplication-factors, the width of the transition-region (measured from signal level 95% to 5% above baseline) is plotted vs. FMAX in Fig.-1b. The width of the transition region decreases rapidly from ≅11.5mm to ≅1.6mm for amplitude-multiplication-factors lower than 6.

Discussion: The measurements show that an FMAX=6 is sufficient to reach a profile-sharpness of ≅1.6mm. With this and the given bandwidth a minimum slab-thickness of 11.7mm is possible. This is sufficient in most cases even for single-slice inversion. Higher FMAX only marginally enhances the profile-sharpness and requires a gradient-strength often too high for thin-slice inversion. References: 1. Ordidge et al.: MRM 36,562-566(1996)

160 Quantitative numerical and experimental analysis of the B1+ and SAR distribution of a loaded 3 T proton surface coil G. Wübbeler, F. Seifert, H. Rinneberg; Biomedical Optics and NMR-Measuring Techniques, Physikalisch-Technische Bundesanstalt, Berlin, GERMANY. Synopsis: Pronounced distortions of the spatial B1-field distribution arise at higher frequencies (f > 100 MHz) in the body. The knowledge of the B1 distribution is important for multichannel receive methods (SENSE, SMASH) as well as for adaptive coil control schemes in the multichannel transmit case. For safety assessment the calculated SAR values have to be calibrated to the input power. In this work the distribution of |B1+| and temperature/SAR were determined experimentally and numerically in for a loaded 3 T proton surface coil. Methods: The measurements were performed on a 3 T whole body scanner (MEDSPEC30/100, Bruker Medical). The coil was loaded either with a spherical or a cylindrical phantom (both: δ = 20 cm, ε = 76, σ = 0.33 S/m). The transmit B1+-distribution was determined by applying a preparation pulse followed by a gradient echo sequence for imaging of the remaining magnetization. From the amplitude dependence of the image intensity the corresponding B1+ value was calculated. The temperature distribution resulting from CW-heating was recorded by MR-thermometry. The numerical calculation were performed with the XFDTD software package (REMCOM Inc). The models were implemented on a 0.5 cm grid. The surface loop was excited by applying a broadband current pulse. From the resulting oscillating and decaying voltage at the feeding port the resonance frequency, the impedance

S85

and the loaded quality factor was determined. The impedance was used to calibrate the calculated field values with respect to the input power. At each location of the simulation grid the complex steady state field values for the electric and magnetic field values were determined and used to calculate the distribution of B1+ and SAR. Results: The absolute values of the simulated |B1+| values, calibrated by the impedance, differed from the experimental values by less than 15 %. After correction of this deviation by scaling the simulated |B1+| values apparently match to a high degree to the experimental values (Fig. 1, |B1+| in µυT/(kW)0.5). The recorded temperature distributions also agreed well the SAR-distribution obtained from the FDTD calculation. Discussion: It follows from the achieved quantitative agreement between measured and numerical values, that the electromagnetic interaction between the surface loop and the load is correctly simulated by our model. Expanding these experimental and numerical methods to phased array antenna configurations will provide helpful information for the development of high field multichannel transmit/receive coil techniques.

161 Actualized evaluation of the relative contributions of intrinsic noise sources in NMR with cryogenic RF probes J. C. Ginefri, M. Poirier-Quinot, L. Darrasse; Unité de Recherche en Résonance Magnétique Médicale, CNRS, UMR 8081, Orsay, FRANCE. Introduction: Recently, many SNR improvements were reported using cryogenic probes i.e. HTS coils or cold copper coils. However significant gains can only be expected when the internal noise of the coil in it’s initial configuration is the dominant loss mechanism, but corresponding application domains have not been fully identified yet. In this work, we investigate the relative contributions of noise sources in NMR when using cryogenic RF probes. The prospect is based on state-of-the-art performances of cryogenic coils with regards to classical scaling laws. Methods: Most significant noise sources are the internal noise of the coil RCTC, the sample noise RSTS and the radiation noise RRTR, R and T being the equivalent resistances and temperatures of the sources. Their contributions were compared in the case of a planar circular coil applied at the surface of a semi-infinite conductive sample. This was done taking into account the coil radius and the NMR frequency. For HTS coils, the effect of static field amplitude was also considered. Scaling laws were used to compute RS [1] and RC for copper coil [2] and HTS coil [3, 4] and normalized with earlier Q-values [3, 4]. RR was calculated using theoretical expression [5]. Results: Limits at which RSTS equal RCTC are shown Fig. 1. Each curves delimitates a lower domain where the internal noise is dominant, indicating that SNR improvement can be achieved by means of temperature reduction or using HTS material. Radiation noise remain generally negligible as compared to the sample noise. However, it must be considered for unloaded coils as

S86

RF-System

shown in Fig.2 using the limits at which RRTR equal RCTC. Discussion and Conclusion: Our actualized noise limits provide a useful guide to determine the best appropriated coil technology when addressing SNR improvements. The Q values of HTS coil are known to be strongly dependant on coil design and technological processes. However internal losses derived from highest reported Q values at very different frequencies are coherent with generally assumed ω2 behavior, corresponding to a linear Q increase as the NMR frequency decreases. 1. Chen CN, Hoult DI [1989] Biomedical Magnetic Resonance Technology. N-Y: Adam Hilger. 2. Hoult DI, Lauterbur PC. [1979] J.Magn.Reson.34: 425-433. 3. Ginefri J.C, Crozat P, Darrasse L [2002] CIMTEC 2002. 3: 449-456 4. Ginefri J.C, Darrasse L, Crozat P [2001] Magn.Res.Med 45: 376-382 5. Kraichman M, [1962] J. Res. N.B.S., D. Radio propagation, 66D, no. 4: 499-503

162 Comparative study of intravascular RF coil designs A. Benattayallah1, J. T. Heverhagen1, V. Matschl2, H. Alfke1, K. J. Klose1, H. J. Wagner1; 1Diagnostic Radiology, University Hospital, Marburg, GERMANY, 2Siemens Medical Solutions, Siemens, Erlangen, GERMANY. Introduction: Several studies showed that intravascular singleloop coils could be used to improve the signal-to-noise ratio (SNR)

and resolution for evaluation of the vessel wall1,2,3. However, SNR in the near field of the dipole, which is the region of interest (ROI) for vessel wall imaging could be improved by innovative coil designs. The purpose of this work was to study the magnetic field homogeneity improvement by using multi-loop for intravascular coil design. Methods: In this study, a computer simulation has been applied to evaluate and optimize the magnetic field and the signal drop for multi-loop coils. The magnetic field is calculated from the given current distribution using Biot-Savart law. In our simulation we varied the number of loops (1,2,3 or 4) and the conductor position to search for an optimised coil performance. The diameter of the coils was fixed to 5 mm. Experimentally, we tested the validity of our simulation by measuring, for each coil, the SNR of a cylindrical phantom containing 1.25g/l CuSo4 solution. All investigations were carried out in a 1 T Magnetom Expert scanner (Siemens, Erlangen, Germany). Results: Using an optimised conductor position for all coils: the single-loop (SL), double-loop (DL), triple-loop (TL) and quadruple-loops (QL) coils, the calculation of the magnetic field distribution at a distance of 3mm from the surface of each coil, result in magnetic field inhomogeneities of 41%, 30%, 18% and 11%, respectively. The dipole character of the single-loop design lets the magnetic field magnitude decay with distance r as 1/r2 . Figure 1 shows the magnetic field distribution for each coil, the higher the number of loops the better the field homogeneities. The measured SNR in MRI images of the phantom confirm these results as shown in Figure 2.

RF-System

S87

Discussion: A single-loop design leads to a significantly higher signal intensity close to the conductors compare to the points that are in the middle between the two wires. This signal variation is reduced as we increased the number of loops. Obtained simulations are in good agreement with experimental results. Coils with fourloops provide a good field homogeneity with the same SNR and penetration depth compared to single-loop coil. Such coils may enable the acquisition of a high resolution MRI images for intravascular wall depiction. References: 1. Quick et al., Magn Reson Med. 41(4): 751-758 (1999). 2. Matschl V et al., VASA 30: 9-13 (2001) 3. Gesine G et al., Circulation,99: 1054-1061 (1999).

163 3 Tesla general purpose transceive helmet coil H. Merkle1, W. Driesel2; 1Lfmi, NINDS/NIH, Bethesda, MD, 2Max Planck Institute of Cognitive Neuroscience, Leipzig, GERMANY. Introduction: When compared to a birdcage head coil, a helmet coil has a better filling factor and improved SNR at the superior part of the brain due to its geometry [1, 2]. The previously reported designs were optimized for high sensitivity at motor cortex, however, the trade-in was a relatively large B1-gradient along the z-axis. In order to also perform fMRI studies in deeper regions of the brain with improved homogeneity and sensitivity we had to modify the original design as described below. Methods: Experimental results were obtained on a 3T Bruker 30/100 MEDSPEC system. The flip-angle distribution was imaged by the double-angle method [3] using a standard multi-slice spinecho imaging sequence along with a cylindrical gel phantom (1.6wt% agarose; 20-cm diameter; 23-cm length). Coil wire arrangement and shapes were altered from a four-fold to a virtually eighth-fold symmetry with additional distributed capacitors. A balanced tune/match network in combination with a quarter-wavelength balun was utilized and a parallel resonance circuit for suppression of standing waves was added on the outer jacket of the semi-rigid cable for additional stability. Results and Discussion: Axial homogeneity is directly comparable to that of the birdcage coil. There is still a moderate B1 gradient along the z-axis, about half of that of the original helmet coil (figure 2). Experiments verify that this coil permits investigations the entire brain with improved SNR factor of up to 1.7 compared to the used birdcage. The modified design provides sufficient space for use of additional audiovisual stimulation devices and appears well suited for both anatomical and functional studies at 3 T. The helmet coil is simple and easily to construct and therefore offers a good alternative to various other transceive head coil designs. Figure 1: Design of the helmet coil Figure 2: Coronal MDEFT images of the human head (256*256 pixels, field-of-view 25 cm, slice thickness 5 mm, TI=650ms) acquired with the standard head birdcage coil (left), with the original helmet coil (middle), and with this design (right). References: 1. Merkle H, et al [1993] 10th Ann. scientific meeting, ESMRMB, Rome, Italy, p.123 2. Merkle H, et al [2000] 8th Ann. meeting of ISMRM, Denver, p. 56 3. Stollberger R, Wach P [1996] MRM 35: 246-251.

164 Resonators for in-vivo and ex-vivo MRI at field strengths up to 17.6T V. C. Behr, I. Wieland, M. Oechsner, D. Gareis, T. Neuberger, D. Haddad, A. Haase, C. Faber; Department of Physics, EP5 (Biophysics), Würzburg, GERMANY. Introduction: Along with increasing field strengths, new challenges for hardware design arise. Resonators designed to operate at frequencies up to 750MHz will be presented. Their performance is demonstrated in in-vivo and ex-vivo imaging experiments. Results of studies on phantoms as well as on preserved dolphin fetuses and living mice are shown. Setup: The Bruker Avance 750MHz system employed in this research is a 17.6T wide bore vertical magnet for which two gradient units are available: a 1T/m system with an inner diameter of 40mm and a 200mT/m system with 56.5mm. Resonator designs optimized for 1H (750MHz), 19F (705.6MHz) and 23Na (198.4MHz) imaging are presented. Hardware: For the 200mT/m gradient system several resonators were built for use in in-vivo studies. Two of the resonators shown in figure 1 offer an inner diameter of 38mm for an animal handling system. While linear operation was readily achieved with birdcage 1a) at a length of 40mm, the system was very sensitive to even small differences in the values of the capacitors employed when operated in quadrature. 1c) is a double-tuned saddle coil setup which was optimized for 23Na imaging of mice and small rats with a proton channel available for positioning and shimming. It is a combination of two perpendicular saddle coils with the larger saddles being tuned to 198.4MHz and the smaller ones to 750MHz. The surface coil shown in figure 2 allows larger samples than the volume resonators described above. Designed for the 40mm gradient system 1d) is a 20mm inner diameter linear birdcage which is 42mm long while 1b) offers 27mm inner diameter in order to hold small animals like mice thus leaving only 7.5mm on either side for socket, circuits and shielding. Experiments and Conclusions: Experiments were performed on phantoms, preserved dolphin fetuses and mice. An axial slice of a living mouse (figure 3) acquired with birdcage 1a) and a sagittal image of a dolphin fetus with resolutions of below (100µm)3 [1]

S88

RF-System

(figure 4) acquired with resonator 1d) are shown. Homogeneity and B1 field strength of all resonators proved sufficient to record images of highest quality. With careful selection of electrical elements and an optimized design, birdcage or saddle coil setups perform well at 17.6T thereby granting access to the benefits of high fields like larger chemical shifts or improved SNR. References: 1. D.Haddad, V.C.Behr, H.H.A.Oelschläger, A.Haase, "High Field 3D-NMR Imaging of Dolphin Embryos and Fetuses", ISMRM Proc., Honolulu, 2002

165 An improved probebase and coil design for in-vivo body MRI of small mammals at 17.6T V. C. Behr, I. Wieland, M. Oechsner, D. Gareis, T. Weber, T. Neuberger, A. Haase, C. Faber; Department of Physics, EP5 (Biophysics), Würzburg, GERMANY. Introduction: For recent high field spectrometers, especially invivo imaging of inner organs of rat-sized mammals requires elaborate hardware to provide as well room for the animal handling system and the animal as for the RF resonator and its feeds and connectors. An improved setup will be introduced in this abstract and the performance will be evaluated in studies on the spinal cord and heart of rats. Setup: Experiments are preformed on a Bruker Avance 17.6T wide bore vertical magnet with a 200mT/m gradient unit for in-vivo studies offering 56.5mm inner diameter. The rats are studied invivo under iso-fluorane anesthetic. Breath and heart beat are monitored with a custom-built monitoring unit that is also employed to trigger experiments in order to avoid motion artifacts. Hardware: An optimized probebase and coil was designed for the study of inner organs of rats. Balancing units (baluns) in the feeds are a pre-requisite especially for advanced designs like quadrature resonators since at frequencies up to 750MHz (larmor frequency of 1H at 17.6T) the electrical setup is very sensitive even to minor imbalances. The probebase shown in figure 1 combines two feeds including baluns with an increased sample volume. To maintain the larger available volume within the resonator a transmit-receive surface coil mounted on a half-cylinder (figure 2) is used. A shield made of copper foil with a thickness of only 7µm to avoid eddy currents on a polyimide substrate covers the outer side of the halfcylinder (figure 3). The animal resting on a slide can be positioned from outside the magnet. This animal handling system is designed to be compatible to already existing commercial systems, so it can be employed in other probebases and/or with volume resonators. The probebase itself consists of a GFP outer cylinder, POM leadthroughs for 2 feeds with baluns and 4 PVC rods for tuning and matching. A PVC/aluminum basis offers two BNC connectors and a fixable adjustment screw for slide-positioning. Experiments and Conclusions: In preliminary studies images on healthy rats were acquired. Figure 4 shows an axial slice with the spinal cord clearly visible and figure 5 is a short-axis view of a rat’s heart. In both cases the new setup offers a sufficiently homogeneous and strong B1 field and yields highest quality results. It provides a successful combination of animal handling capabilities, large sample volume and electrical demands for advanced resonator design for MR microscopy.

Clinical

Young Investigators Finals Clinical 4:00 pm - 5:00 pm

S89

Willem Burger

166 Automatic segmentation and plaque characterization in atherosclerotic carotid artery MR images I. M. Adame1, R. J. van der Geest1, M. Mohamed2, B. A. Wasserman2, J. H. C. Reiber1, B. P. F. Lelieveldt1; 1LKEBRadiology, LUMC, Leiden, NETHERLANDS, 2Radiology, Neuroradiology Div., Johns Hopkins Hospital, Baltimore, MD. Introduction: Composition and structure of atherosclerotic plaque is a primary focus of cardiovascular research. High resolution carotid artery magnetic resonance imaging (MRI) has a potential role in stroke prevention by identifying vulnerable plaque. To quantitatively assess the degree of vulnerability and the type of plaque, the contours of the lumen, outer boundary of the vessel wall and plaque components need to be traced. This is usually performed manually, which is time-consuming and sensitive to interand intra-observer variability. Our goal is to develop an automated contour detection technique that requires as little manual interaction as possible. Methods: The algorithm consists of three different phases: Outer boundary of the vessel wall: ellipse-fitting approach. In this work, an ellipse centered in the middle point (provided by the user) is deformed, rotated and mapped to the vessel, according to gradient measurements. The ellipse with the greater gradient average will be taken as the outer boundary. Afterwards, a minimum cost approach (based on dynamic programming) is used to refine the

S90

Clinical

contour. Lumen: fuzzy clustering approach. Lumen segmentation is performed using a classification based on the pixel gray value followed by a minimum cost approach (similar to that for the outer boundary). Information from the first phase is also taken into account, since the lumen is constrained within the outer vessel wall boundary. Plaque: fuzzy clustering approach: same approach as for lumen but plaque is more difficult to segment as the pixel gray value can differ considerably from one region of plaque to another, even when it corresponds to the same tissue. Information from lumen and outer boundaries is used to constrain plaque detection to the area within the vessel wall. Results: The algorithm has been optimized for 50 high resolution PD and T1 weighted MR images (17 patients) (pixel size 0.54 mm). Results: Excellent correspondence between automatic and manual area measurements: lumen (r=0.92), outer (r=0.91); and acceptable for fibrous cap thickness (r=0.71). The average paired difference between the automatic/manual measurement pairs was -2±5.2mm2 (9.1±21.7%; p<0.05) for lumen area, showing a slight area underestimation. The outer contour area error was -5.6±19.5 mm2 (5.4±19.1%; p=0.05), and fibrous cap thickness error -0.1 ± 0.6 mm (7.8 ± 48.6%; p = NS); these errors were non-significant. Conclusions: Though further optimization is required, our algorithm is a powerful tool for automatic detection of lumen and outer boundaries, and characterization of plaque in atherosclerotic vessels.

167 In vivo MR imaging of brain inflammation in human ischemic stroke A. Saleh1, M. Schroeter2, H. Hartung2, U. Moedder1, S. Jander2; 1Institute of Diagnostic Radiology, University Hospital Duesseldorf, Duesseldorf, GERMANY, 2Department of Neurology, University Hospital Duesseldorf, Duesseldorf, GERMANY. Introduction: Brain infarction is accompanied by strong inflammation in the ischemic tissue area. Histopathologically, macrophages constitute the vast majority of inflammatory cell populations in the infarcted brain. They produce a wide spectrum of potentially cytotoxic mediators and may contribute to secondary exacerbation of ischemic brain injury. We designed a clinical phase II pilot study to test the suitability of USPIO-enhanced MRI for the noninvasive visualization of macrophage influx in human ischemic stroke lesions. Subjects and Methods: This study was performed on 10 consecutive neurologic inpatients. Within 24 hours after symptom onset, all patients received stroke imaging revealing acute ischemic lesions. Major exclusion criteria were cerebral hemorrhage, clot lysis with alteplase, or enrollment to other clinical studies. Study design was approved by local ethics committee. USPIO (AMI-227, Sinerem®) was a kind gift of Laboratoire Guerbet (Roissy, France). To define the pattern of conventional contrast medium uptake a first follow-up MRI was performed 4 to 5 days after symptom onset. This was followed by a single USPIO infusion (2.6 mg of iron per kilogramm of body weight) USPIO 24 - 36 hours later, with two additional follow-up MRIs 24 -36 and 48 -72 hours after USPIO infusion. All MR examinations were performed with a 1.5T magnet (Vision; Siemens) by using conventional T1-weighted and T2-weighted spin-echo sequences and T2*-weighted gradient echo sequences. Diffusion- and perfusion weighted imaging was perfomed with single-shot echo-planar sequences. Gadoliniumenhanced images were obtained 4 – 5 days after stroke and 48 - 72 hours after USPIO infusion. Results: USPIO infusions were well-tolerated by all patients. In T1-weighted MRI, USPIO lead to hyperintense signals predominantly marginal but also within the infarct zone. In all patients receiving a Gadolinium enhanced T1 scan before USPIO infusions, the pattern of gadolinium enhancement greatly differed from that obtained upon subsequent USPIO adminstration. This distribution is compatible with macrophage accumulation demonstrated by immunohistochemistry of experimental brain infarcts and human autopsy material. Discussion: USPIO enhanced MRI renders new pathophysiological information in subacute human stroke. In all patients studied patterns of USPIO-related T1 signal alterations could not be retrieved by any other signature of multimodal MRI or a conventional contrast agent. Data from experimental stroke models strongly suggest that USPIO-induced T1 hyperintensity indeed reflects macrophage accumulation in ischemic stroke lesions. Therefore, USPIO enhanced MRI may provide a in vivo surrogate marker for the development of inflammation-targeted therapies in stroke and other CNS pathologies.

Clinical 168 Benign Prostate hyperplasia – Evaluation of treatment response with DCE-MRI J. T. Heverhagen1, H. von Tengg-Koblik1, K. T. Baudendistel1, G. Jia1, H. Polzer1, H. Henry1, A. L. Levine2, T. J. Rosol2, M. V. Knopp1; 1Department of Radiology, The Ohio State University, Columbus, OH, 2Department of Veterinary Biosciences, The Ohio State University, Columbus, OH. Purpose: To demonstrate the changes after treatment of benign prostate hyperplasia (BPH) and evaluate therapy success. An experimental study in beagles with a naturally enlarged prostate was carried out. Subjects and Methods: 12 male beagles (mean age: 4.4 ± 0.9 years) were investigated in our study. These animals develop BPH naturally in an early age. All animals were investigated five times, twice before and three times during a treatment with an anti 5alpha reductase treatment. Investigations were carried in four week intervals. The treatment commenced immediately after the second examination. The animals were divided into two groups of six dogs each; one group was treated with the actual drug and the second with a placebo. The dogs were placed in the head coil of a 1.5 T clinical scanner (Twinspeed, GE, Milwaukee, WI). Besides three dimensional volume measurements in T1- and T2-weighted sequences, DCE MRI measurements were carried out using a threedimensional fast SPGR sequence (TR: 7.6ms, TE: 2.6ms, FA: 25°, 28 slices, Acquisition time: 24s, 32 time points). After three acquisitions serving as a baseline, 0.4 mL/kg bodyweight contrast agent (Prohance, Bracco, Princeton, NY) were injected with an injection rate of 0.2 mL/s. Subsequently, a flush of 15mL saline solution followed. A two compartment model was used to fit the data on a pixel by pixel basis. Results: While the treatment resulted in a volume reduction of more than 50%, the prostate volume in the control remained unchanged. Moreover, the pharmaco-dynamic analysis showed additional distinct differences between both groups. Prior to treatment, the gland could be divided into an outer, peripheral zone and an inner, periurethral zone. The periurethral zone presented a higher vascularity demonstrated by a steeper signal increase and longer retention of the contrast agent. During treatment, the periurethral zone broadened and the vascularity of the outer zone adapted the features of the inner zone. After three months, the two zones were no longer distinguishable in the treated dogs. In comparison, no change of vascularity was observed in the control dogs. Conclusion: DCE MRI allows to depict changes in vascularity of the prostate due to treatment of benign prostate hyperplasia. In addition to volume measurements, this provides different parameters for the evaluation of therapy response. Figure 1: Investigation of a dog prior and two months after treatment. The evaluation curves show the differences in vascularity of the periurethral zone after treatment.

S91

Science

S92

Young Investigators Finals Science 4:00 pm - 5:00 pm

Jurriaanse

169 3He lung ventilation imaging with B1-inhomogeneity correction in a single breath-holdscan W. Miller, J. P. Mugler, III, T. A. Altes, M. Salerno, E. E. de Lange, J. R. Brookeman; Department of Radiology, University of Virginia, Charlottesville, VA. Purpose: Hyperpolarized 3He MR images are typically acquired using a close-fitting transmit/receive surface coil. B1-field inhomogeneity is therefore significant in some portions of the lung and leads to artifactual intensity variations in ventilation images. These artifacts vary with patient, due to differences in coil fit and loading, and can mimic disease. Artifacts may be differentiated from real ventilation defects by measuring and correcting for the RF-coil sensitivity distribution. The sensitivity map would ideally be obtained in the same breath hold as the ventilation image. We demonstrate a method to simultaneously acquire sensitvity maps and ventilation images during a single breath hold, and compensate for intensity variation artifacts. Methods: Our pulse sequence collects two interleaved, 5/8 partialFourier images per slice, both at the same nominal flip angle θ, in opposite phase encoding directions. The first interleave covers central k-space near the end of the scan, while the second covers it near the beginning. The available signal declines throughout the scan as (cosθ)n-1, where n is the excitation number. Since the voxel magnitude is dominated by information from central k-space, each interleaved partial-Fourier image has a different flip-angle weighting, and the ratio of the two images is used to extract a flip angle map. A single ventilation image is created from both interleaves by adding common phase-encoding lines and performing partialFourier reconstruction on the combined data set. Finally, each image voxel is multiplied by a correction factor based on the local flip angle. Three normal volunteers and one lung transplant recipient were imaged using both our technique and a standard FLASH sequence on a 1.5T scanner (Siemens Sonata). All scans used the same vest-shaped RF coil. The bandwidth was set 50% higher for our B1-correction technique to keep imaging time equal for both sequences. Results: B1-inhomogeneity corrected images consistently showed less severe intensity variation artifacts than corresponding FLASH images, but had 20% less SNR due to higher bandwidth. The figure shows a set of coronal images from a normal volunteer. The FLASH image shows several possible ventilation defects. The B1corrected image indicates that some of these are artifactual.

Conclusion: We demonstrated the ability to simultaneously obtain coil sensitivity maps and ventilation images during a single breathhold 3He scan, and use the sensitivity maps to compensate for B1field inhomogeneity artifacts. The resulting images show the potential for improved diagnostic quality over conventional images, but have slightly decreased SNR for equal imaging time.

170 Vasomodulation of skeletal muscle BOLD signal D. P. Bulte1,2, G. Duench1, O. McGregor1, J. Alfonsi1, M. D. Noseworthy2,1; 1Diagnostic Imaging, The Hospital for Sick Children, Toronto, ON, CANADA, 2Department of Medical Biophysics, University of Toronto, Toronto, ON, CANADA. Purpose/Introduction: Substances which affect brain blood flow are known to cause perturbations in brain BOLD signal. However, effects of these vasomodulators on muscle BOLD signal have not been extensively investigated. Previously we have demonstrated that oscillating subject breathing of 100% O2 with normoxia can produce a ‘boxcar-like’ pattern in muscle BOLD signal similar to that seen in fMRI brain activation studies; the amplitude of which is likely related to voxel blood perfusion and volume. Using oxygen-enhanced BOLD imaging of human calf muscles, we proposed that muscle fibre differences (fast twitch vs. slow twitch) would be observable and that caffeine and antihistamines, known vasomodulators, could result in differences between pre and post exercise states. Subjects and Methods: The right calf in male subjects was immobilized and imaged using a GE Signa 1.5T CV/i MRI and transmitreceive imaging coil. A gradient echo sequence (10 slices 6mm thick, 16cm FOV, TE/TR=40/1500ms, 2NEX, α=90°) with a spiral readout was employed to collect 180 temporal points, during which the subjects periodically inhaled 100% O2 (4 cycles of 90s normoxia each followed by 45s of 100% O2). Subjects then exercised (200 rapid calf-raises) and the imaging protocol was repeated. For studies to evaluate vasomodulators fasted (8 hr) subjects were similarly evaluated, however the imaging/exercise paradigm was repeated 1 hour following ingestion of either caffeine (100 mg) or 2 hours following antihistamine (4mg chlorpheniramine). Muscle BOLD dynamic range, ∆BM (%BOLD signal difference, post vs. pre-exercise) was calculated and compared between slow and fast twitch muscles. Results: Hyperoxia was effective as a modulator of muscle BOLD signal (fig.1). ∆BM was greatest in soleus. Caffeine had no effect on soleus ∆BM, however, chlorpheniramine significantly dimished this exercise response (Table 1). Gastrocnemius muscle showed lower basal ∆BM, relative to soleus, and opposite to the vasomodulation effect seen in soleus, caffeine resulted in a dramatic increase in ∆BM while chlorpheniramine showed no effect. Table 1 No Drug

Caffeine

Antihistamine

Soleus (fast twitch)

11.27±2.34

11.81±3.86

2.65±0.49

Gastrocnemius (slow twitch)

7.21±2.76

25.42±7.51

6.44±3.97

Molecular Imaging: MR or PET?

S93

information on the time-scale of oxygen transfer in various lung diseases. References: 1. Muller CJ, et al [2002] Radiology Feb; 222(2): 499-506 . 2. Deichmann R, et al [1999] MRM 42:206-209. 3. Jakob PM, et al [2001] JMRI 14:795-799.

Fig.1. Pre-exercise functional map (left) compared to the same leg post exercise. Discussion/Conclusion: Exercise induced blood flow and volume increases can be studied using hyperoxia-modulated BOLD imaging. These results suggest chlorpheniramine (antihistamine) could act detrimentally on fast twitch muscle performance, while caffeine could improve slow twitch muscle performance.

171 Imaging lung function by dynamic acquisition of T1-maps J. F. T. Arnold, T. Wang, E. D. Pracht, P. M. Jakob; Department of Physics, University of Wuerzburg, Wuerzburg, GERMANY. Purpose/Introduction: Recently the feasibility of imaging lung function with oxygen-enhanced MRI was shown [1]. Relative enhancement ratios and mean slopes of relative enhancement were measured with T1-weighted sequences. As an alternative, we report a technique to dynamically image lung function with oxygen-enhanced MRI using T1-parameter-maps, which allows an accurate, quantitative assessment of the mean slope of T1-enhancement and therefore lung function. Because T1 mapping reveals information about a tissue specific parameter, dynamic T1 mapping is more accurate than working with T1-weighted sequences, in which the results depend on the exact data acquisition parameters. The possibility of fast and dynamic T1 mapping, without waiting for full relaxation between the acquisition of maps, has already been reported [2]. Based on this work we developed an even faster method of measuring T1 dynamically, with higher SNR than reported in [2]. Subjects and Methods: Three healthy volunteers were examined on a clinical 1.5 whole body scanner (Vision, Siemens, Erlangen) with a maximum gradient strength of 25 mT/m and a slew rate of 83 T/m/s. For signal detection a 4-element body phased array coil was used for signal reception. Lung T1-maps, based on an IR Snapshot FLASH technique [3], were acquired. The imaging parameters for the Snapshot FLASH technique were TE=1.0ms, TR=3.5 ms, FA=7°, matrix=64*128, FOV=500 mm2 and slice thickness=15 mm. One T1-map was acquired every 6.7 seconds. Results: Table 1 confirms that our dynamic T1-procedure provides correct T1 values compared to standard T1 measurements. Figure 1 shows a oxygen enhanced dynamic T1-map acquisition in a volunteer. The curve shows the mean T1 values of the upper right lung. The slope of oxygen wash-out is calculated to be -1.32 ± 0.41 1/T1/msec. The slope of oxygen wash-in is 3,57 ± 0,88 1/T1/msec, about three times higher in magnitude. Discussion/Conclusions: With oxygen enhanced dynamic acquisition of T1-maps slopes of T1 enhancement can be measured. Thus, a new parameter for assessing lung function is available. Therefore, the proposed method has the potential to provide quantitative

Hot Topic Debate Molecular Imaging: MR or PET? 5:15 pm - 6:15 pm

Willem Burger

172 In favour of MR C. Moonen; Laboratory for Molecular and Functional Imaging, University of Bordeaux 2, Bordeaux, FRANCE. MRI has several major advantages over PET with respect to molecular imaging: 1. Versatility: Images can be obtained depicting anatomy as well as physiology (perfusion, diffusion, flow, motion) to specific pathway processes with or without contrast agents 2. No ionizing radiation is used, and MRI follow-up of molecular therapies can be carried out without harmful effects 3. Detailed anatomical information is always obtained in addition to molecular imaging setting the stage for image-guided therapy 4. PET imaging is much slower than MRI. In addition, the total temporal window of PET imaging depends on the lifetime of the specific radiolabel, whereas MR contrast agent generally maintain their function so long as their physiological environment remains identical. As a result, MRI can be used more easily for tracking migrating cells (e.g. stem cells) during prolonged periods of times 5. MRI systems are cheaper, and more widely available than PET systems. In addition, PET systems require a nearby cyclotron. MRI contrast agents are cheaper and do not need to be produced locally.

S94

Cardiac Imaging: MRI vs. CT

MRI has one major disadvantage as compared to PET: its insensitivity. Without contrast agents, MRI can image compounds in the millimolar range, whereas PET can measure concentrations in the nano-or even picomolar range. Therefore, MRI requires amplification strategies for specific pathway labeling. As demonstrated by Meade et al., this can be achieved by nanoparticles filled with contrast agent. The particle wall was designed to be digested by a locally expressed enzyme, and the liberation of a large quantity of contrast agent then generated the amplification effect. In addition, macrophages, lymphocytes and stem cells have been shown to take up large amounts of USPIOs without modifying basic functions. Much research is needed to achieve this amplification for all major biochemical pathways. MRI has already replaced PET for functional brain studies, and once molecular MR probes have been developed, it might do so for heart as well. Image guided molecular therapy (local drug delivery, gene therapy, local drug activation) is part of molecular imaging and has a tremendous future. PET alone may suffer because of the lack of detailed anatomic information. The combination PET/CT is helpful but increases the exposure to ionizing radiation. In addition, CT has a very limited role in molecular imaging. Therefore, in order to make PET a very versatile technique in molecular therapy of the future, a combination PET/MRI should be developed.

173 In favour of PET G. K. von Schulthess; Medical Radiology, University Hospital, Zurich, SWITZERLAND. Nuclear Medicine (NM) and in particular Positron Emission Tomography (PET) represent the only imaging modalities today, which have and had extensive capabilities to image function including molecular events in a clinical environment. There is only one reason for this. Tracer amounts in the nano- to picomolar range can be detected with high efficiency, while all the other imaging modalities are at least a factor of 10’000 less sensitive to “contrast” enhancement. For this reason, nuclear techniques have had much success in providing examinations ranging from the assessment of perfusion to imaging of brain receptor system functions. Development of tracer molecules for NM and PET have taught us, that “small is beautiful”. Small tracer molecules have favourable kinetics to reach the tissues where they are targeted to. With this high sensitivity, the small biomolecules – sometimes minimally modified for labelling purposes – can serve as the molecular probes. PET with Fluorodeoxyglucose (FDG) in particular has developed rapidly in the last few years into a major clinical tumor staging modality, but FDG-PET has applications in early therapy control and inflammation imaging as well. Images obtained are a survey of extended body areas, which can be obtained in 10 to 30 minutes with the newest imaging devices. In particular, the advent of PET/CT has added speed, fast quantification and anatomic reference to PET images, which makes them relatively easily interpretable and very frequently provides the decisive imaging “answer” required. Thus, many suggest, that PET/CT is the future one-stop-shop oncology examination. However, other applications of PET are also important, namely brain and cardiac imaging, where there exists substantial information, that quantification of data is relatively easily achieved. In summary, NM and particularly PET have properties, which make it currently very hard to imagine that they can be matched by

other function oriented imaging techniques in the near future, particularly when thinking of widespread clinical applications.

Hot Topic Debate Cardiac Imaging: MRI vs. CT 5:15 pm - 6:15 pm

Jurriaanse

174 In favour of MRI D. Pennell; CMR Unit, Royal Brompton Hospital, London, UNITED KINGDOM. CMR is an ideal modality for investigating the heart because of several key attributes: 1. Freedom from ionising radiation 2. Non-toxic contrast media 3. Acquisition in any plane without the need for reformatting 4. Versatility of available physiological information 5. Extensive clinical validation in numerous clinical and research reports These issues will be discussed during this debate.

175 In favour of CT V. Sinitsyn; Tomography Dept., Cardiology Center, Moscow, RUSSIAN FEDERATION. For many years use of CT in cardiac imaging had been quite limited. Only dedicated electron-beam (EBCT) scanners have been capable to provide artefact-free images of heart chambers and coronaries without motion artefacts. Situation has changed dramatically after 1998, when multislice CT (MSCT) scanners appeared on the market. For the first time in the history of radiology the general-purpose CT system could be applied for sophisticated cardiac imaging. Major advantages of MSCT and EBCT are fast volumetric scanning, possibilities for prospective or retrospective ECG-synchronisation, high spatial resolution and ability to detect calcifications. Even 4-spiral MSCT with 0.5s tube rotation allow one to obtain excellent images in most cardiac patients. MSCT systems with 16spirals having 0.4s tube rotation times and submillimeter spatial resolution give even better results. Temporal resolution of modern CT scanners could be as low as 80-125 ms per scan for multislice systems and 30-50 ms - for EBCT. Complete volumetric study of the heart and coronaries can be done just in one breath-hold (1835s). Besides that, modern cardiac CT gives high-resolution details of the heart chambers morphology including the cardiac valves. Precise measurements of functional parameters (ejection fraction, volumes, myocardial thickening) can be done with cardiac CT. New generation of EBCT imagers has excellent possibilities for quantitative studies of myocardial perfusion, contractility and blood flow. Comparative studies with MRI had shown that coronary CT angiography in general gives better results (higher sensitivity, specificity and better overall accuracy - Ropers D et al [2002] Invest.Radiol. 37:396-392, van Geuns et al [2002] Am.J.Cardiol 90:58-63). Cardiac CT is faster, more robust and has better spatial resolution. It can be regarded as a best non-invasive modality for coronary imaging. Of course, cardiac CT, being x-ray technique,

Cardiac Imaging: MRI vs. CT requires proper justification for use and optimisation of radiation exposure. Situation in cardiac imaging is very competitive and a choice of the diagnostic modality strongly depends on its availability and impact on patient's diagnosis and management. Cardiac MSCT and EBCT are widely used for some of the most important medical and social indications like screening of coronary atherosclerosis (calcium scoring) and non-invasive coronary imaging. Taking into consideration progress of cardiac CT, we cannot any more consider MRI as a sole “one-stop-shop” for cardiac imaging. Each diagnostic modality should be used according to its possibilities and real clinical need for such type of studies. Fig. 1. MSCT of stented LAD. Fig. 2. Perfusion defect (arrow) seen with EBCT.

S95

S96

MR Angiography and MR Contrast Agents II / Small Animal Imaging and Spectroscopy II

Saturday, September 20, 2003

Mini-Categorical Course MR Angiography and MR Contrast Agents II 8:00 am - 9:00 am

Willem Burger

176 Contrast-enhanced MRA of the aortic arch and tributaries V. Martinez de Vega; Resonancia Magnetica, Clinica Ntra. Sra. del Rosario, Madrid, SPAIN. Contrast-enhanced magnetic resonance angiography (CEMRA) is increasingly becoming a useful non-invasive imaging method for the assessment of thoracic aorta. The technique is fast, does not require EKG gating and provides a 3D data set which can be reformatted into any desired plane. In opposition to TC and conventional angiography, the acquisition of CEMRA images does not involve ionizing radiation nor potentially nephrotoxic contrast agents are required, both things being of great advantage in the evaluation of young patients. Most MR strategies for imaging the aorta include CEMRA, to assess vessel lumen, combined with black blood spin echo sequence for depiction of the aortic wall, and in selected cases, a phase-contrast sequence for the evaluation of functional flow dynamics. In this lecture technical aspects of CEMRA will be described. Current and potential clinical applications in aortic arch pathology will be reviewed as well. Several important issues to optimize image quality, like contrast dose, injection rate, scan delay time will be analyzed. Recent improvements in gradient design, together with the use of new strategies for K-space scanning (Time Resolved Imaging Contrast Kinetics –TRICKS-, Radial Projection reconstruction) have all resulted in ultrafast imaging. Hence the possibility of acquiring multiple 3D volume set every few seconds within a single breath hold is already a reality. CEMRA is a very useful tool to evaluate developmental anomalies including aortic coarctation, double aortic arch, aberrant left subclavian artery, aberrant right subclavian artery and right side aortic arch. Other pathologies affecting the aortic arch and its branches like aneurysm, stenoses and dissection can be easily assessed with MR. CEMRA can be used for the non-invasive postoperative follow-up, so making it a method of choice for children.

177 Contrast-enhanced MRA of the coronary vasculature S. Dymarkowski; Radiology, UZ Gasthuisberg, Leuven, BELGIUM. As intravascular contrast media prove their potential to selectively enhance vessels in different areas of the body, excitement is building among MR users, since this decreases one of MRI most important weaknesses as an angiographic tool: the need to sacrifice spatial resolution for imaging time. Intravascular agents, also known as blood pool agents, offer a distinct advantage over conventional MRI contrast agents. Rather than diffusing quickly into the interstitial space, they remain in the blood for extended periods of time. As a result, it is possible to

acquire steady-state images over an hour or more, rather than chasing a conventional contrast bolus for at maximum 25 to 30 seconds. Though no one can say with certainty, many investigators are predicting that intravascular contrast will play a key role in finally capturing the most sought-after target in noninvasive angiographic imaging : the coronary arteries. Blood pool contrast agents work by a variety of mechanisms. Ironbased products, take advantage of their large molecular size, which prevents diffusion into body tissues. Gadolinium-based agents stay in the blood stream as a result of transient binding to albumin. Albumin binding offers an additional benefit beyond localization in the blood pool. The contrast agent begins to spin much more slowly, at the rate albumin spins, causing a relaxivity gain that produces a substantially brighter signal than would be possible with freely circulating gadolinium. Recent studies have shown that qualitative assessment of the visualization of coronary arteries and their side branches was significantly better with use of blood pool agents than with use of extracellular contrast agents, and also when compared to nonenhanced state-of-the-art MR coronary angiography techniques, such as sequences with T2-preparation.. The majority of blood pool agents are at this moment still in the experimental phases. Once they are tested clinically, they may allow rest-stress myocardial perfusion imaging in addition to coronary artery imaging in a single imaging session. Other approaches besides blood pool agent research are currently also being extensively studied. These include the use of steady state free precession techniques and arterial spin tagging. These techniques are valuable to visualize coronary arteries by using the natural high contrast of blood and selective excitation of inflowing blood respectively. Further studies are needed to determine which technique will prove most valuable in the future.

Mini-Categorical Course Small Animal Imaging and Spectroscopy II 8:00 am - 9:00 am

Jurriaanse

178 Anaesthesia and physiological monitoring M. G. Sommers; Central Animal Laboratory, University Medical Centre Nijmegen, Nijmegen, NETHERLANDS. In most of the animal studies using magnetic resonance imaging and/or spectroscopy, anesthesia is a necessary part of the experimental protocol. However, the outcome of the experiments can be seriously influenced directly by the properties of the used drugs or indirectly by the disturbance of the animal's physiology and homeostasis. Therefore, it is advisable to choose an anesthesia protocol with care and pay some attention to minimizing the side effects of the chosen anesthetics by additional measures like thermoregulation, providing supplementary oxygen, maintaining fluid balance and artificial ventilation. To keep a good impression of the animal's status, additional MR compatible anesthetic monitoring devices are needed to estimate the efficacy of the above measures, thereby providing the feedback that is needed to adjust the anesthesia equipment. In conclusion it can be said that the use of well chosen, optimized anesthesia protocols will improve the relevance of the results found in animal models and facilitate the extrapolation to humans.

MRI in Oncology: Diagnosis and Staging II 179 Hardware aspects (RF and gradient coils, the use of clinical scanners for high-res MRI, ultra high-fields, etc) C. Faber; Biophysik, EP5, Uni Wuerzburg, Wuerzburg, GERMANY. Compared to humans, MRI of small animal models requires enhanced resolution with more than a 1000-fold smaller voxels to visualize the corresponding structures. The main parameters accessible to improve the performance of NMR are the strength of the main magnetic field, the strength of the magnetic field gradients, and the size and design of the high-frequency resonators. Clinical scanners can be enhanced by the use of high-performance gradient systems and dedicated animal resonators. One can also take advantage of the large bore size to examine several animals simultaneously in a multiple-coil setup. Dedicated animal scanners, commonly working at a field-strength up to 7 T, provide higher sensitivity. Typical bore sizes up to 40 cm allow for convenient animal handling while at the same time providing optimized gradient systems and resonator designs. The use of quadrature resonators or the combination of surface-receive and volume transmit resonators allows one to reach microscopic resolutions. The highest sensitivity can be achieved with ultra-high-field magnets with field strengths up to 17.6 T. However, the small bore sizes (typically 89 mm) strongly restricts the design of gradient systems, animal handling systems, and high-frequency coils. Particular problems in resonator design arise from the reduced wavelengths, for example small capacitor values, electric field couplings, and dielectric resonances. Some of the latest developments which overcome these problems will be presented.

Mini-Categorical Course MRI in Oncology: Diagnosis and Staging II 8:00 am - 9:00 am

Ruys

180 Soft tissue tumours J. Galant; Radiology, Hospital Universitario San Juan de Alicante, Alicante, SPAIN. Staging: No universal accepted staging system for soft tissue sarcomas exists. Grade should clearly be weighted heavily in any staging system as well as the extent and size that directly influence the prognosis. On the other hand, the primary factors governing the type of local treatment to be used are the compartment involved and the proximity of the tumor to vital neurovascular and osseous structures. To achieve the local control, surgery should be preferably radical or wide. Radiologist should aid the orthopaedic surgeon while planning the surgery providing all the information regarding extension to the neurovascular structures, bones, joints, muscles and fascial planes. There is no a universal protocol to assess the distant extent of sarcomas. As the lungs and bones are the most common sites for disseminated disease, they should be routinely studied preferably by CT and scintigraphy. Those sarcomas with predilection for lymphatic spread are best evaluated by pelvic and abdominal CT if they involve the lower extremities. In addition, the tendency of myxoid

S97

liposarcomas to spread the soft tissues makes recommended to rule out retroperitoneal and intra abdominal involvement. Characterization: Presurgical knowledge of the histologic nature of a soft-tissue tumor is of utmost importance in the preoperative planning. As experience with MR imaging of the soft tissue tumors has been increasing, reliance in the capability of characterization of these lesions has also been growing. Our approach to the problem of characterization of a soft-tissue mass begins looking for a specific diagnosis that can be strongly suspected in many cases such as lipoma, hemangioma, pigmented villonodular sinovitis, nerve sheath tumors, giant cell tumors of tendon sheath, elastofibroma and fibromatosis. In some other cases the diagnosis of pseudotumor as hematoma, fat necrosis, elastofibroma or infection can be made, as well as with some malignant tumors. When a specific diagnosis is not achieved, characteristics of benignity and malignity should be analysed. There are many variables to be studied, especially relationships of the tumor with the tissues in vicinity. Diagnosis is also based on signal intensity, textural patterns and epidemiological variables. Likewise, patterns of enhancement in dynamic studies can be of some utility limiting the differential diagnosis. To manage such amount of data, we also employ an automatic diagnostic tool based on a neural network. Probably the most remarkable feature of the management of a soft tissue mass is that communication between surgeons and radiologists should be as close as possible.

181 Pancreatic malignancies W. Schima, MD; Department Radiology, University of Vienna, Vienna, AUSTRIA. Introduction: Pancreatic cancer is the fourth or fifth leading cancer death in the Western world. The most common form is ductal adenocarcinoma, which carries a poor prognosis (5-year survival <5%). At the time diagnosis, radical resection of the tumor, which offers the only chance for cure nowadays, is feasible in only 20-30% of the patients. The major task of radiologic studies is to reliably diagnose or exclude pancreatic cancer in patients with suspected pancreatic masses. In patients with pancreatic cancer, appropriate staging regarding the presence or absence of vascular invasion, lymph node involvement, liver metastases, and peritoneal seeding should be performed. Materials and Methods: It is important to use high-field-strength MR units (* 1.0 T) with strong gradients and a phased-array torso coil, which makes breath-hold imaging possible in order to avoid motion artifacts. T1-weighted GRE images (FLASH, Siemens; TFE, Philips; SPGR, General Electric) with or without fat saturation and a slice thickness of not more than 5 mm are used. For T2weighted imaging TSE pulse sequences are preferable. T2-weighted TSE sequences with fat saturation (for detection of liver metastases) and without fat saturation (for delineation of the pancreas) are employed. In case of suspected pancreatic cancer one set of T1-weighted and T2-weghted images each should cover the pancreas and liver in order to provide complete staging. MRCP pulse sequences (HASTE) yield additional information for pancreatic head tumors. As contrast agents either gadolinium agents or mangafodipir trisodium are used. Dynamic gadolinium-enhanced T1-weighted 3D-GRE images reveal better delineation of the vasculature to evaluate vessel infiltration by tumor. However, mangafodipir-enhanced MRI has been found to be superior to

S98

Perfusion Imaging: heart-brain axis

single-slice CT in the detection of small cancers and in the delineation of liver metastases. Discussion/Conclusion: With the advent of multi-slice CT, CT imaging has gained increasing popularity for pancreatic imaging because of ist exquisite 3D-visualization techniques of the peripancreatic vessels. However, MR imaging is still superior in the detection of liver metastases and in the delineation of small pancreatic tumors. Due to its superior soft tissue contrast, MR imaging is also the method of choice in the differential diagnosis between tumors and tumor-simulating conditions in patients with equivocal CT.

Plenary Session Perfusion Imaging: heart-brain axis 9:20 am - 10:50 am

Willem Burger

182 Perfusion MRI of the brain: techniques, indications and findings P. Parizel, J. W. Van Goethem, Ö. Özsarlak, A. M. De Schepper; Department of Radiology, Universitair Ziekenhuis Antwerpen (University of Antwerp), Antwerp, BELGIUM. Together with the heart, the brain is one of the most intensely perfused organs of the human body. Though the brain represents only 1/50th of body weight (1400 g), it receives approximately 15~20% of the cardiac output. Overall cerebral perfusion is in the order of 50-60 ml/min/100 g tissue weight. Perfusion is higher in the metabolically more active gray matter (70-80 ml/min/100 g tissue) than in white matter (30-45 ml/min/100 g tissue). Increasingly, imaging techniques such as MRI and CT can be used to elucidate the pathophysiology of ischemic stroke.the purpose of perfusion imaging is to show the ischemic penumbra, which is a potentially salvageable region of decreased cbf, in which the threshold for irreversible cell death has not yet been reached. in patients with (hyper)acute stroke, intravenous thrombolysis with recombinant tissue plasminogen activator is of proven benefit within 3 hours of symptom onset, and intraarterial thrombolytic therapy shows promise within a 6 hour window of therapeutic opportunity. On MRI, perfusion weighted imaging (PWI) can be performed using two basic approaches. in arterial spin labeling (ASL), hydrogen protons are labelled outside the head and the flow of this endogenous contrast agent (i.e. tagged spins) through the brain is observed. This technique is not yet widely used in clinical practice, because it suffers from poor snr and necessitates long imaging times. in this presentation, we shall focus on the currently more widespread approach which uses injection of an exogenous contrast agent (Gd-chelate) to act as a T2* contrast agent during its first pass through the cerebral vasculature. the contrast agent causes a transient decrease in signal intensity (T2*-shortening susceptibility effect), proportional to the concentration in a given region. The technique is known as dynamic susceptbility contrast (DSC) imaging. Using a rapid imaging sequence (typically EPI), as many as 50 sequential images can be obtained during bolus injection of contrast, covering a time interval of roughly 70 seconds. Bolus injection of contrast is performed 5 to 10 seconds after the start of the imaging sequence, to ensure that an adequate number of baseline images are obtained.We inject a contrast volume of 0,2 mmol/kg body weight (i.e. 30 ml for a 75 kg person) at an injection rate of 5 ml/sec (antecubital vein, 18-gauge IV catheter). This is followed by injection of 20-30 ml of saline, to flush the gadolinium

out of the tubing, arm vein and lung vasculature. When available, a power injector should be used, although with some experience, adequate results can also be obtained with hand injection (two syringes, containing contrast and saline flush, connected to a bifurcated “y” check valve system). Sequential images are acquired simultaneously in multiple slice positions. The EPI sequence that we currently use performs 50 chronological images in 12 slice positions, yielding a total dataset of 600 images. If more slice positions should be required, this would necessitate a longer TR, and this would negatively influence the quality of the sequence. Perfusion images can then be processed in parameter maps such as: time to peak (TTP), percentage of baseline (PBL), mean transit time (MTT), cerebral blood volume (CBV), cerebral blood flow (CBF), which can be calculated using the formula: CBF = CBV / MTT. The parametric perfusion images are then compared to the diffusion-weighted scans, in order to determine the extent of the ischemic penumbra. In summary PWI has improved our understanding of the pathophysiology of cerebral circulation, and is likely to become an important clinical tool for selecting (hyper)acute stroke patients which are potential candidates for thrombolytic therapy.

183 Perfusion MRI of the heart: techniques, indications and findings M. Jerosch-Herold; Department of Radiology, University of Minnesota, Minneapolis, MN. The application of contrast-enhanced, ultra-fast MRI to assess myocardial blood flow is a promising method for the detection of coronary artery disease and ischemic cardiomyopathies. The quantitative determination of a myocardial perfusion reserve with MRI has shown good concordance with invasive tests, such as the measurement of the coronary flow reserve. The severity of an epicardial coronary stenosis, assessed by quantitative coronary angiography in patients with chest pain symptoms, was in good agreement with the hyperemic response measured with MR perfusion imaging. Myocardial blood flow measured with MRI in a large population of asymptomatic men and women was found to be age and gender dependent, heterogeneous, and it correlated with the cardiac workload at rest. The perfusion reserve changed significantly with risk score, indicating that myocardial perfusion imaging with MRI may play a useful role in assessing the impact of risk factors and detecting subclinical disease. Furthermore, new parameters such as the ratio of endo to epi-cardial perfusion, measured only at rest, may under some circumstances enhance the sensitivity for detecting disease. The MRI techniques most prevalent for the assessment of myocardial perfusion are based on ultra-fast, T1weighted gradient echo sequences. For such a technique, we have previously validated the quantification of absolute myocardial blood flow, both with extracellular and intravascular Gd-contrast agents. The upper and lower 95 % confidence limits for the relative differences in myocardial blood measured with MRI and labeled microspheres were found to be +/- 40% in a recent experimental study in our laboratory. Absolute myocardial blood flows (MBF) measured with MRI in healthy normals have similar characteristics to MBF’s measured in previous studies with quantitative PET in a population of asymptomatic subjects. Further optimization of the contrast-to-noise ratio and the image acquisition speed promise to yield significant improvements upon the results achieved to-date. (Work supported by R01 HL65394-01 and R01 HL65580-01 from NIH/ NHLBI)

Diffusion Imaging: Methodology 184 Determinants of myocardial perfusion evaluated quantitatively by EBT R. Rienmüller; Abteilung für Radiologische Diagnostik, Universitätsklinikum Graz, Graz, AUSTRIA. Purpose/Introduction: The essential physiological mechanisms that regulate coronary and myocardial blood flow are: physical factors as cardiac output, aortic pressure, vascular and myocardial resistance. they are controlled by • intrinsic auto-regulation • metabolic demands • neural and humoral stimulation or blockade Subjects and Methods: 1722 patients with known or suspected coronary heart disease were studied by ebt using standard protocol to measure myocardial perfusion in ml/100g/min and to determine functional lv-parameters and to evaluate coronary calcification score. Heart rate and blood pressure was controlled in all patients. Results: The mean myocardial blood flow in male and female was 64 ± 18 ml/100g/min (n=984 male) and 78 ± 20 ml/100g/min (n=524 female). using histogram classification of myocardial blood flow, 5 different classes were found for both sex. there was no direct correlation between the extend of coronary calcification, the severity of coronary stenotic lesions and global myocardial perfusion. Myocardial perfusion was closely related to the heart rate, to the cardiac output and to the double-product of heart rate and blood pressure. Myocardial blood flow may be in normal range in spite of several stenotic lesions above 50%, when tachycardia and/or high blood pressure are present; may decrease without stenotic lesions above 50% when bradycardia and/or low cardiac output and/or low blood pressure are present. Discussion/Conclusion: Myocardial perfusion maybe measured on routine daily basis using ebt. it is for the first time that under real clinical conditions it is possible to confirm the importance of the knowledge of the physical factors which determine the actual myocardial blood flow being under the control of auto-regulative, metabolic and neural und humoral stimulation or blockade.

185 Similarities and differences between cerebral and cardiac perfusion M. R. Rees; Radiology, University of Bristol, Bristol, UNITED KINGDOM. The normal cerebral blood flow is 45-50ml/100gm/min ranging from 20 ml in white matter to 70 ml in grey matter. The brain receives about 2% of the cardiac output. And maintains a mean arterial pressure independent of perfusion pressure unless there is severe hypotension. The heart itself receives approximately 5% of the cardiac output and receives a flow of 200-250 mls./min at rest. Coronary flow varies according to the time point in the cardiac cycle and is regulated to respond to exercise rather than maintain a steady state. Coronary flow reserve is a measure of the hearts ability to respond to stress, this stress response can be induced pharmacologically in cardiac perfusion imaging which has as its aim the measurement of differences between stress and rest perfusion. Stress perfusion imaging has traditionally been carried out in the heart by using isotopic methods, with agents such as technetium or thallium. Cerebral stress perfusion imaging is also possible using isotopic methods with agents such as Diamox which can enhance

S99

perfusion by 30-50% this method is particularly useful in cerebral PET imaging. Chronic ischaemia in the heart produces a state of myocardial hibernation. It has become important to establish this diagnosis as patients with this diagnosis may benefit from revascularisation where patients with myocardial scarring will not. Until recently the most sensitive method of diagnosis of cardiac hibernation has been carried out with PET scanning using FDG. This type of scanning is also particularly sensitive in demonstrating causes of myocardial ischaemia, mental stress has been shown to be a significant factor in producing cardiac ischaemia in addition to physical stress. New developments in traditional radiological imaging techniques are now being used in the heart to assess myocardial perfusion as well as anatomy and physiology following on from the pioneering work in the brain that has already been carried out. Electron beam CT has been used for many years to measure myocardial blood flow the use of multislice CT in myocardial perfusion measurement is currently under experimentation. Myocardial perfusion imaging by MRI has been established and combining first pass techniques with late perfusion imaging has been shown to be a promising method of detecting and quantifying ischaemic myocardium and separating viable myocardium from scarred tissue. "Happiness comes only when we push our brains and hearts to the farthest reaches of which we are capable." -Leo C. Rosten

Scientific Session Diffusion Imaging: Methodology 11:20 am - 1:00 pm

Willem Burger

186 Direct implementation of diffusion tensor imaging and f-MRI data in stereotactic radiosurgery B. Stieltjes1, B. Didinger2, M. Amann3, T. Wilhelm4, M. Essig1; 1Radiology, German Cancer Research Institute, Heidelberg, GERMANY, 2Radiotherapy, German Cancer Research Institute, Heidelberg, GERMANY, 3Physics, German Cancer Research Institute, Heidelberg, GERMANY, 4Neuroradiology, Universtaetsklinikum Heidelberg, Heidelberg, GERMANY. Introduction: Stereotactic radiosurgery enables dose reduction in viable structures adjacent to the target volume like functional cortical areas and nerve fibres. f-MRI and Diffusion Tensor Imaging can delineate such structures . Here we report our first experience with importation and co-registration of such MRI data in a radiotherapy planning system (RPS). Our goal is to spare these viable area's and reduce radiotherapy complications. Subjects and Methods: In an ongoing study, three patients with cerebral arterio-venous malformations, presenting with word finding problems or partial hemiparesis, were included. f-MRI at1.5 T. Block design paradigm for identification of Broca´s and Wernicke´s area (1). Scan parameters: TR/TE/α: 4s/54ms/90°; 22 slices, 4 mm thickness, 128 x 128 matrix, FOV of 240 x 240 mm; 90 volumes, four and a half off/on cycles. Motion correction (2) and functional analysis using AFNI (3). Diffusion sensitised Single Shot EPI at 1.5 T. Scan parameters: TR/TE: 180/107; 55 slices, 2.5 mm thickness; 128x128 matrix FOV 256 x 256 mm; 6 non-collinear diffusion gradients, 2 b-values 0 and 1000; 6 averages. Calculation of fractional anisotropy and color coded images (4). MRI data transfer to the RPS, realignment to conventional

S100

Diffusion Imaging: Methodology

MRI and CT, image alignment check, final target delineation and dose delivery depiction. Results: Fig. 1 A shows Wernicke's area. Significant voxels were used to create a mask shown in B. This mask was matched on CT (C) and used for calculation of the treatment plan. 80 % isodosis outlined in magenta, Wernicke's area, oulined in yellow, could be spared.

Fig. 2 A Colormap . The white arrow in fig. 2 A indicates the right capsula interna, the red arrow indicates the deformed left capsula interna containing fibres responsible for residual right-sided motor function. AVM depicted on T2-weighted images (red arrow in fig 2 C), residual fibres cannot be delineated. The cross hair in Fig 2 B-C indicates the position of the mouse in one image and the matching pixel in the realigned image. As can be appreciated, matching is accurate.

Discussion: It is possible to accurately spare viable structures depicted using f-MRI and DTI. Follow-up is needed to evaluate if sparing of these viable area's leads to decreased complications of radiosurgery. References: 1. Hampson M. et al.[2002] Human Brain Mapping 15:247-262. 2. Thesen S. et al. [2000] Magn. Reson. Med. 44:457-465. 3. Cox RW. [1996] Comput. Biomed. Res. 29:162-173. 4. Pajevic S. et al. [1999] Magn. Reson. Med. 42(3):526-540.

187 Reproducibility of diffusion tensor imaging between two different MR scanners E. J. Vlieger, E. M. Akkerman; Radiology, Academic Medical Center, Amsterdam, NETHERLANDS. Purpose: To determine the reproducibility of Diffusion Tensor Imaging (DTI) between two different MR scanners. To our best knowledge, this has never been reported before Methods and Materials: DTI was measured at 16 well-defined para-axial slice locations in the brain, in 11 healthy male volunteers, and on two scanners. The scanners were a General Electric Signa Horizon and a Siemens Vision, both 1.5 T and both equipped with EPI gradients. Diffusion weighted (DW) acquisitions were performed in 6 icosahedric directions (Akkerman EM, MRM 2003) using modifications of the standard single shot DW-EPI sequences of the manufacturers. A 3D T1-weighted scan was acquired as a reference for white/grey matter. To correct for Eddy-current-induced geometric distortions, each DW image was matched to the corresponding slice without DW, by a 2D affine transformation. Per scanner, the average voxel-displacement was calculated.

Fractional anisotropy (FA) images were calculated both from the uncorrected and the corrected image set. White/grey matter segmentation was performed in the 3D data set, and the FA images were matched to the 3D data set. Finally, for each subject, the ratio of the average FA in white matter between the 2 scanners was calculated. Results: The average voxel displacement was 1,75 ± 1.91 mm for the Vision and 2,39 ± 1.44 mm for the Signa (p < 0.001). The average ratio of FA values in white matter was (Signa/Siemens) 0.91 ± 0.07, both for the geometry-uncorrected and the geometry-corrected FA images. This differed significantly from 1.0 (p < 0.001) Conclusion: The Signa suffered more from eddy current induced distortions than the Vision. This is most probably caused by the fact that on the Vision a double echo sequence was used, which was not yet available on the Signa. The FA values of both scanners do not reproduce well: the white matter FA values from the Signa are 9 percent lower on average than the values from the Vision. We have as yet no clue as to the cause of this discrepancy.

188 Effects of artefacts correction on diffusion tensor imaging at 3 T E. Cassol1, J. Lätt2, E. Larsson3, S. Brockstedt3; 1Biophysics and multimodality Imaging - Neuroradiology MRI, University Hospitals, Toulouse, FRANCE, 2Department of Radiation Physics, Lund University, Lund, SWEDEN, 3Department of Radiology, Lund University Hospital, Lund, SWEDEN. Introduction: Diffusion tensor imaging (DTI) provides information about tissue integrity through measurements of changes in water mobility. Accurate quantification of the mean diffusivity (trace) and fractional anisotropy (FA) suffers however, from various artefacts. Our aim was to compare the accuracy of DTI measurements after correction using either an iterative method (1) or a registration method based on mutual information (SPM99). Subjects and Methods: Six healthy volunteers and five patients with multiple sclerosis underwent DTI examinations with (FlairDTI) and without an inversion pulse, respectively. Images were acquired using a 3.0 T Siemens Magnetom Allegra head scanner. Scanning parameters: DTI: b=0, 300, 1000 s/mm2 with diffusion encoding in 6 directions ((1,01), (-1,0,1), (0,1,1), (0,1,-1), (1,10), (-1,1,0)); TE/TR=131ms/4000ms, bandwidth 752 Hz/pixel, matrix 128×128, FOV 220mm. 17 5mm-thick transversal slices (interslice gap 0.5mm), were positioned parallel to the bi-commissural line. For b=1000 s/mm2, 5 measurements were obtained in order to improve the signal-to-noise ratio. Flair-DTI: TR=9000ms, TI=2500 ms. The other parameters were identical. Image post-processing was performed using an in-house developed IDL-based diffusion analysis program (Research Systems, Inc.). For all volunteers, ROIs were positioned according to their localization on the un-weighted images, in the frontal white matter and in the corpus callosum. For patients, additional ROIs were located in lesions. Results: Visual inspection of all images revealed small qualitative differences (Fig). In healthy volunteers, the (STD/measurement)ratio (%) was increased on corrected images for both indices. With IR pulse, this ratio is highly reduced for all the methods. Moreover, the IR-pulse contributes to reduce the artefacts occurring due to CSF in the frontal part of the brain. However, in MS lesions, this difference is reduced for the trace while for FA, the ratio is increased on images with IR. Discussion/Conclusion: The images acquired in this study were less affected by geometric distortions than expected. The best

Diffusion Imaging: Methodology correction was obtained with the iterative method, which, however, requires the acquisition of images with an intermediate b-value. The registration method, widely used in MS studies (2), is not well adapted to correct for geometric distortions occurring due to eddy currents, since it is based on rigid body transformations, but it gave good results with limited interpolation. Finally, the use of an inversion pulse did not significantly improve the results of the correction but it increased the confidence in the measurements. References: 1. de Crespigny et al ISMRM; 1998:661(2) Cercignani et al AJNR; 2001:22 (952-58)

189 Classification of bundles of white matter tract from DTI without registration P. Batchelor1, D. Atkinson1, A. Connelly2, F. Calamante2, D. Tournier2, D. L. G. Hill1; 1Imaging Sciences, King's College London, London, UNITED KINGDOM, 2Institute of Child Health, University College London, London, UNITED KINGDOM. Purpose/Introduction: We propose here a particular use of mathematical tools from the geometry of curves in order to quantify fibre tracts from DT-MRI independently of their spatial position (see also Basser(1997) Ding(2003)). We also extend the geometry to the relative spatial configurations of curves, describing for example how spatial curveswind around each other, using quantitities like the Link of a pair of curves (Eppie(1998))(inspired from DNA and polymer folding analyses Fain(1999)). We apply this method to in vivo datasets.

S101

Methods-Theory: The fundamental theorem of space curves states that space curves are uniquely classified by their curvature and torsion and can be reconstructed by integration of the Frenet-Serret equations. We define a distance function on tracts shape as a distance function on curvatures and torions of the curve normalised to a constant length. The Link of a pair of tracts is defined as an integral similar to Bio-Savart law. MR diffusion datas from two healthy volunteers were used. Acquisition: 1.5T Siemens Vision system (20 directions, 60 slices twice same volunteer ; 7 directions, 25 slices, 6 averages) Results: Intra-dataset comparison Figure 1 a) and b) show an example of the main application, the comparison of mean shapes between two in vivo datasets, here dataset 1 and dataset 2, in this specific case the bundles of fibres were chosen arbitrarily. This was done, as announced, without any need for registration, which illustrates the usefulness of the concept. Inter-dataset comparison: intra-subject and inter-subject In this experiment, we computed some fibre tracts in the Corpus Collosum, and display their curvatures and torsion. The main result is in Fig. 2, showing one of the tracts in the dataset 1 , and the closest in shape in dataset a second dataset. The local Link between two tracts is shown in Figure 3. Discussion-Conclusion: We have demonstrated techniques that allow the classification and comparison of fibre shapes. We have also introduced a method to study the relative shapes of multiple fibres, and in particular to quantify their spatial relationship using the link. In summary, the use of the Frenet-Serret equations seems a particularly natural way of normalising shapes. References: 1. Basser(1997) Annals of the NYAS 820, p 123-138. 2. Ding (2003) Magn.Res.Med., 49, p 716-721 3. Fain (1999) Phys. Rev.E 7239-7252 4. Eppie (1998) Math. Intell. 20 p45--52

190 High b-value DTI of the brain- a novel contrast A. Zaman1, U. Wieshmann2, N. Roberts1; 1Magnetic Resonance and Image Analysis Research Centre (MARIARC), University of Liverpool, Liverpool, UNITED KINGDOM, 2Neuroradiology, The Walton Centre for Neurology and Neurousurgery (WCNN), Liverpool, UNITED KINGDOM. Purpose: MRI imaging allows the detection of structural abnormalities in patients with neurological condition such as stroke, MS or epilepsy. However, a large number of patients have entirely normal standard MRI. For example, 70% of all patients with partial epilepsy have normal MRI scans despite suffering from frequent seizures and having focal electroencephalographic changes and histopathological abnormalities. There is therefore a need for the development of novel image contrast techniques which are able to

S102

Diffusion Imaging: Methodology

detect the underlying abnormalities in such patients to improve the yield of MRI. The development of diffusion imaging and diffusion tensor imaging (DTI) has offered the possibility to go beyond anatomical imaging and study tissue structure at a microscopic level in vivo. The purpose of the present study was to explore the capability of DTI to produce novel contrast at high b-values. Materials and Methods: DTI was performed using a 1.5T GE LX system (maximum amplitude 40 mTm-1). The DTI pulse sequence we employed is an adaptation of a diffusion weighted (DW) single shot EPI enabling 25 diffusion directions with b values of 1000, 2000, and 3000 s mm-2. Results: Figure 1 shows apparent diffusion coefficient (ADC), fractional anisotropy (FA), and b max maps for b of 1000, 2000, and 3000 s mm-2. The difference between b1000 and b3000 is zero in CSF (as expected because CSF has like water a mono exponential diffusion characteristics). There are however marked differences in comparing b values of 1000 and 3000 s mm-2 in the brain indicating a deviation from the mono exponential diffusion characteristics in brain tissue (ADC maps in Figure1). White matter, in particular in the posterior part of the brain show a more prominent deviation from the mono exponential diffusion characteristics than grey matter (b max maps in Figure1). Discussions: Our results reveal a non-monoexponential signal decay in normal brain tissue. This effect was more prominent in posterior white matter than in grey matter or anterior white matter. The non-monexponential diffusion characteristics provide a novel contrast. At the present the underlying microstructural mechanisms causing the diffusion characteristics are poorly understood. However, it is very likely that any change to the microstructural environment will affect the diffusion characteristics and therefore result in a detectable alteration of the contrast on high b value diffusion maps. We are currently applying DTI with high b values in patients with epilepsy with and without structural abnormalities on standard MRI.

191 Salivary glands in healthy volunteers: comparison of apparent diffusion coefficients (ADC) before and after stimulation H. C. Thoeny, F. De Keyzer, F. G. Claus, S. Sunaert, R. Hermans; Radiology, University Hospitals Leuven, Leuven, BELGIUM. Purpose: To evaluate whether diffusion-weighted Echo Planar Imaging (DW – EPI) of the salivary glands can demonstrate functional changes during and after salivary stimulation. Subjects and Methods: Twelve healthy volunteers (5 women, 7 men) with a median age of 25 years (range: 22-30) were examined on a 1.5T MR unit (Sonata, Siemens, Erlangen, Germany). Axial T1-w SE and T2-w TSE-sequences were acquired for morphological evaluation of the salivary glands. Thereafter a diffusion-weighted EPI-sequence using four b factors (b = 400, 600, 800 and 1000 sec/mm2) was obtained with the following parameters: slice thickness 3.5mm, intersection gap 0.7mm, and FOV 230x230mm. The total acquisition time of this sequence was 2:22 min. All sequences were performed once in basal circumstances, and only the DW EPI was repeated during and after salivary stimulation with one tablet of 500mg ascorbic acid given orally over a mean period of 26 minutes. For statistical analysis Student's t-tests were performed. Results: In basal circumstances, the apparent diffusion coefficient (ADC), calculated from the DWI sequences, was significantly lower in the parotid glands (0.857+0.011 mm²/sec) than in the submandibular glands (1.326+0.016 mm²/sec), p < 0.001. During the first 5 minutes of salivary stimulation, a significant decrease in ADC was observed (p<0.05) in all volunteers, both in the parotid as well as submandibular glands. During the following 10-15 min a steady increase to a significantly higher value compared to baseline (p<0.05) was noted for the parotid glands, while the ADC of the submandibular glands returned to baseline (p=0.89). The ADC value of the parotid glands did not completely recover by the end of the examination. Conclusion: Diffusion-weighted echoplanar MRI allows to noninvasively demonstrate functional changes in the salivary glands.

192 In vivo visualization and quantitative analysis of diffusion in the human spinal cord S. Kwiecinski1, P. E. Summers1, H. Mamata2, S. E. Maier2, S. S. Kollias1; 1Institute of Neuroradiology, University Hospital Zurich, Zurich, SWITZERLAND, 2Department of Radiology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA. Introduction: The purpose of this research is to use the diffusion weighted line scan MRI to obtain information about the microscopic organization of the human spinal cord in vivo. Subjects and Methods: Diffusion-weighted images of the spinal cord were acquired in the axial and sagittal planes in healthy volunteers on a 1.5-T clinical MR system using the line scan diffusion imaging (LSDI) method [1] with an in-plane resolution of 0.7 - 1.0 mm and a slice thickness of 5 mm. Diffusion weighting with a bfactor of 750 s/mm2 was applied in 6 directions. Scan time per slice was 15/30 min. Images were reconstructed off-line to produce 1st and 2nd eigenvector direction maps as well as Fractional Anisotropy (FA) and Apparent Diffusion Coefficient (ADC). The FA and ADC maps were quantitatively analyzed for the entire cord cross section and grey and white matter compartments. Results: Good quality in vivo diffusion-weighted images (DWI),

Diffusion Imaging: Methodology and calculated ADC and FA maps of the cervical and lumbar human spinal cord were obtained in the axial plane. The 1st eigenvector maps demonstrate organized, longitudinal directionality consistent with the rostrocaudal arrangement of the ascending and descending white matter tracts. The 2nd eigenvector maps primarily show transverse components consistent with the expected distribution of collateral nerve fibers. (Fig 1,2). Quantitative analysis of the FA and ADC in grey and white matter of the cervical spine segment (C5 level -axial plane) on 7 healthy volunteers is summarized in Table 1. Discussion: Diffusion-weighted imaging and mapping the orientation in space of white matter fibers of the human spinal cord in vivo is feasible in a clinical MR scanner. The quantitative results of a ADC and FA values calculated for the selected ROI can play a significant role in estimating the extent of Spinal Cord Injury. Further studies including thoracic and lumbar crossections in axial and sagittal planes are to be continued. References: 1. Gudbjartsson H et al. [1996] MRM 36: 509-519 2. Robertson R.L. et al. [2000] AJNR 21: 1344-1348 3. Mamata H et al. [2002] ISMRM 10: 544 4. Bammer R et al. [2003] AJNR 24: 5-12

Cervical Spine Diffusion Attributes Diffuon Attributes

Grey Matter

White Matter

FA [%]

49 ± 8

55 ± 7

Entire Spinal Cord 52 ± 5

ADC [m2/s]

(7.7 ± 1.2) x 10-10

(8.4 ± 1.7) x 10-10

(8.5 ± 1.8) x 10-10

193 DTI of the human cervical spinal cord: an application to traumatic injury A. Jasinski1, T. Banasik1, P. W. Stroman2, T. Skórka1, M. Hartel3, M. Konopka3, P. Pieniazek3, B. Tomanek2, A. Urbanik4; 1Radiospectroscopy and MRI, H.Niewodniczanski Institue of Nuclear Physics, Krakow, POLAND, 2MR Technology, Institute for Biodiagnostics, Winnipeg, MB, CANADA, 3Mri, Silesian Diagnostic Imaging Centre Helimed, Katowice, POLAND, 4Radiology, Jagiellonin University Medical College, Krakow, POLAND. Purpose: To determine reference values of the apparent diffusion tensor (ADT) for gray and white matter in the cervical spinal cord (CSC) from the level C2/C3 to C6 for b values up to 1200 s/mm2 using EPI-DTI sequence and to evaluate its application to detection and diagnose of traumatic spinal cord injury in patients with neurological disorders.

S103

Materials and Methods: This study was conducted on 16 healthy volunteers aged from 23 to 63, average 32 and on 16 patients aged from 20 to 74, average 49, with neurological disorders ranging from mild to severe symptoms, following traumatic CSC injury of varying severity. All patients underwent full routine MRI scan of the CSC before DTI scan. All volunteers underwent a basic high resolution sagittal FSE scan before DTI measurements. Imaging was performed on a GE SIGNA LX Echo-Plus, equipped with Research Mode. Standard DW EPI supplied by GEMS was modified to measure diffusion in 6 directions. All DTI scans were gated using a peripheral pulse trigger. Axial DT images were acquired with a 64 x 64 matrix, FOV=7 cm, slice thickness =5mm, slice separation =2mm, number of slices 8 or 10, TR=2 RR or 4 RR, NEX =8 and b factor 300-1200 s/mm2. A single b value DTI scan took around 6min. Values of ADT were determined for different ROI in the GM and WM using an IDL based software developed in-house. Results: Diagnostic quality DWI axial images of the CSC of volunteers and patients were obtained. For b values up to 1200 s/mm2, the principal diffusivities λ1, λ2, λ3 were determined for right and left anterior horns in the GM and for anterior and posterior funiculus in the WM, from C2/C3 to C6 levels. Average values for the GM are: λ1=(0.64±0.08)10-3mm2/s, λ2=(0.65±0.08)10-3mm2/s, λ3=(1.38±0.10)10-3mm2/s and for the WM are: λ1=(0.37±0.08)10-3mm2/s, λ2=(0.38±0.08)10-3mm2/s, λ3=(2.20±0.07)10-3mm2/s (P=0.05). Fig. 1 shows fractional anisotropy indexes FA for GM and WM from C2/C3 to C6. The average values of FA are 0.45 and 0.78 respectively. Fig. 2 shows axial DW images at b=600s/mm2 of a volunteer and a patient after a severe traumatic CSC injury. A region of decreased diffusion is well defined on the right image. In cases of mild neurological disorders DTI did not show any new details, however FA analysis indicated damaged region of the CSC. Conclusion: We have demonstrated the feasibility of clinically practical diffusion imaging of the CSC after severe traumatic injury.

S104

Brain MRS: Cancer, Psychiatry and Neurodegeneration

Scientific Session Brain MRS: Cancer, Psychiatry and Neurodegeneration 11:20 am - 1:00 pm

Jurriaanse

194 Metabolic tumor imaging and absolute quantification of metabolic changes in gliomas A. Stadlbauer1,2, O. Ganslandt1, S. Gruber2, C. Nimsky1, R. Fahlbusch1, E. Moser2,3; 1Department of Neurosurgery, Neurocenter, University of Erlangen-Nuremberg, Erlangen, GERMANY, 2NMR Group, Department of Medical Physics, University of Vienna, Vienna, AUSTRIA, 3Department of Radiodiagnostics, General Hospital of Vienna, Vienna, AUSTRIA. Introduction: Proton magnetic resonance spectroscopic imaging (1H-MRSI) is a noninvasive tool for investigating the spatial distribution of metabolic changes in brain lesions. Brain tumors show increased levels of choline-containing compounds (Cho) and a reduction in N-acetyl-aspartate (NAA) and creatine (Cr). The range of Cho increase and NAA decrease is compatible with the range of tumor infiltration. It is often difficult to delineate the heterogeneous structure of gliomas in conventional MRI. Knowledge of the size of the border zone between tumor and healthy tissue is one of the major problems in therapy planning. Metabolite maps of NAA and Cho allow the differentiation of areas of necrosis, solid tumor and varying degrees of tumor infiltration and tissue edema. Methods: 10 patients (40±11years, 6 male, 4 female), all with supratentorial astrocytomas (WHO grade II and III), and 10 sexand age-matched healthy controls with normal brain MRI were examined. All studies were performed on a 1.5Tesla clinical whole body scanner (Siemens Sonata) using high resolution MRSI (TR/TE =1600/135ms, 24x24 circular phase-encoding, 16x16cm FOV, slice thickness 10mm). Metabolic maps for Cho, NAA and Cho/NAA ratios (Fig.1B) were calculated and a “healthy region” of predominantly white matter (Fig.1A and B, white rectangle) was selected in contralateral brain, at sufficient distance from the lesion to allow segmentation based on the assumption of Gaussian distribution of the Cho/NAA values for normal brain (tested independently). The contours of the segmented tumor (Fig.1C) were used to select the tumor-containing voxels. Absolute metabolite concentrations [Cho], [Cr] and [NAA+NAAG] were calculated using LCModel (Fig.1D and E). Results: In all spectra of patients and controls, we obtained SNR≥3, FWHM≤0.075ppm and SD<20% for 0.4cm3 voxels. As compared to the “healthy region” (contralateral normal brain (CNB)) and to controls, tumor spectra showed significantly higher (t-test for paired samples) absolute concentrations of [Cho] and lower of [Cr] and [NAA+NAAG], respectively (Tab.1). Discussion: For delineation of the border zone of the tumor images of Cho/NAA-ratios are more suitable. The contrast tumor vs. normal brain is improved, because of the utilization of both, the increase of Cho and decrease and NAA. In addition, absolute quantification of MRSI data to study metabolic changes in the whole tumor enables the spatial distribution of tumor infiltration in different patients to be compared. The lower limits for [Cho]=1.44mM and the upper limits for [tNAA]=5.31mM in tumors are used as threshold values for delineation of pathologic changes.

195 Pediatric Pilomyxoid Astrocytomas: MRI and Proton MRSI findings A. Arslanoglu, M.D.1, B. Cirak, M.D.2, A. Horska, Ph.D.1, P. B. Barker, D.Phil.1, T. Tihan, M.D.3, L. Aronson, R.N.4, A. M. Avellino, M.D.5, P. C. Burger, M.D.6, D. M. Yousem, M.D.1; 1Radiology, Johns Hopkins University, Baltimore, MD, 2Neurosurgery, Johns Hopkins University, Baltimore, MD, 3Pathology, University of California, San Francisco, CA, 4Pediatric Oncology, Johns Hopkins University, Baltimore, MD, 5Neurosurgery, Washington University, Seattle, WA, 6Pathology, Johns Hopkins University, Baltimore, MD. Purpose/Introduction: Pilomyxoid astrocytoma (PMA) is a recently described astrocytic tumor with unique histopathologic and clinical characteristics. The most pertinent features of PMA that differentiates it from typical Pilocytic Astrocytoma (PA) are; higher rate of recurrence and CSF dissemination . The purpose of our study was to evaluate the role of MRI and proton MR spectroscopic imaging (MRSI) in preoperative assessment of PMA to take an appropriate therapeutic algorithm. Subjects and Methods: We examined four patients with the diagnosis of PMA (3M, 5 months to 2 years old, mean age at the presentation was 15.8 months). Patients presented with failure to thrive (2 cases), feeding difficulties and developmental delay. Routine brain and spinal MRI with and without GdDTPA administration and proton MRSI of brain were acquired at 1.5 T (GE, Signa, Milwaukee, WI). MRSI was performed using a multi-slice spin-echo sequence with TR/TE 2300/280 ms . Choline (Cho), creatine (Cr), N-acetyl aspartate (NAA) and Lac concentrations were estimated using the phantom replacement methodology. Metabolite ratios Cho/Cr, NAA/Cr, and NAA/Cho were calculated from uncorrected signal areas. Johns Hopkins Medicine Institutional Review Board approved the study. Results: We found predominantly solid tumor, homogenous enhancement, hydrocephalus, extension of T2 signal abnormality into the deep white mater and gray matter, and CSF dissemination

Brain MRS: Cancer, Psychiatry and Neurodegeneration on MRI. MRSI of PMA revealed decreased concentrations of total Cho, Cr, and NAA. In contrast, proton MR spectra of PA showed elevated Cho and decreased Cr and NAA signals. Examples of MRI and MRSI data are shown in the figures.

Fig. 1: Suprasellar pilomyxoid astrocytoma in a 2 year old boy. The lesion is hyperintense on FSE (1a) image. Arrows show extension of high T2 signal into the deep gray/white matter. The lesion shows a homogenous enhancement except central necrosis area (1b). Fig. 2: Proton MR spectra from selected regions of interest in a 2 year old girl. The spectra of the tumor demonstrate low levels of all metabolites and elevated lactate. NAA is mildly decreased in the region adjacent to the tumor. Discussion/Conclusion: Our data show that PMA has different MRI/MRSI characteristics than PA, most likely due to different histopathologic features of PMA. Combined MRI/MRSI may help to distinguish preoperatively PMA from PA. However, larger series are needed in order to be able to use these findings in the differential diagnosis from typical PA.

196 MR spectroscopy guided brain tumor resection: Integration of MRSI in functional neuronavigation using a MRI/MRSI hybrid dataset A. Stadlbauer1,2, O. Ganslandt1, S. Gruber2, C. Nimsky1, R. Fahlbusch1, E. Moser2,3; 1Department of Neurosurgery, Neurocenter, University of Erlangen-Nuremberg, Erlangen, GERMANY, 2NMR Group, Department of Medical Physics, University of Vienna, Vienna, AUSTRIA, 3Department of Radiodiagnostics, General Hospital of Vienna, Vienna, AUSTRIA. Introduction: The main goal of neurosurgery of lesions is to achieve resection as complete as possible while preserving normal brain tissue and function. Information about the localization and spatial extent of the lesion should be available before and while the surgical procedure is performed. Frameless stereotactic methods in neurosurgery (neuronavigation) have been used as a platform to integrate functional information in the operating room, leading to so called functional neuronavigation. In conventional MRI it is often difficult to delineate the heterogeneous structure of lesions, especially of gliomas. Proton magnetic resonance spectroscopic imaging (1H-MRSI) is a noninvasive tool for investigating the spatial distribution of metabolic changes in brain lesions. Subejects and Methods: 14 patients (7 male, 7 female, 45 ± 14 years old), all with untreated supratentorial gliomas (12 with astrocytomas, WHO grades II and III; two with glioblastomas) were ex-

S105

amined on a 1.5 T clinical whole body scanner (MAGNETOM Sonata, Siemens Erlangen, Germany) using high-resolution 1HMRSI (TR/TE 1600/135 ms, 24 × 24 circular phase-encoding, 16 × 16 cm FOV and 10 mm slice thickness). Additionally an axial spin echo (SE) sequence (T1-weighted, TR/TE 500/15 ms, 256 × 256 matrix size, 16 × 16 cm FOV, 20 slices with a distance factor of 0 % and 2 mm slice thickness) was performed. The MRSI slab was precisely aligned to a selected SE slice by copy and paste of the image position. Metabolic maps (256 × 256 matrix size) of Cho, NAA and Cho/NAA ratios were calculated. The MRI/MRSI hybrid dataset was obtained by replacing the raw data of five anatomical slices in the SE data set, correlated with the MRSI slab, with the corresponding Cho/NAA ratio map of the tumor (Fig.1). This hybrid dataset was transferred to a neuronavigation system (VectorVisionSky, BrainLab, Heimstetten, Germany) and registered to a 3D MRI dataset (MPRAGE) resulting in a functional 3D navigational dataset for MRS-guided neurosurgical planning (Fig.2). A navigation microscope (NC4-Multivision, Zeiss, Oberkochen, Germany) enables integrated microscope-based neuronavigation. Results: 12 patients underwent surgery with functional neuronavigation after integration of 1H-MRSI. For two patients surgery is planned at a later time. In all cases, projection of MRSI data into the operating viewing field allowed easy identification of the tumor border based on biochemical information. Conclusion: Functional neuronavigation with integration of 1HMRSI allows presurgical identification of tumor border zones based on the metabolic changes due to tumor infiltration.

S106

Brain MRS: Cancer, Psychiatry and Neurodegeneration

197 Diagnosis of gliomatosis cerebri in tumor patients by proton MR spectroscopy D. Galanaud1, F. Nicoli1, O. Chinot2, S. Confort-Gouny1, Y. Le Fur1, J. Ranjeva1, P. Viout1, D. Figarella-Branger3, P. J. Cozzone1; 1Centre de Résonance Magnétique Biologique et Médicale, UMR CNRS 6612, Marseille, FRANCE, 2Department of Neurosurgery, Hôpital de la Timone, Marseille, FRANCE, 3Department of Neuropathology, Hôpital de la Timone, Marseille, FRANCE. Introduction: Gliomatosis cerebri (GC) is a rare entity characterized by widespread brain tumoral infiltration which lacks objective and quantitative diagnostic criteria. Indeed, WHO’s classification of GC (diffuse glial tumor infiltrating the brain extensively, involving at least two brain lobes) is unsatisfactory, as many low grade gliomas (LGG) involve at least two brain lobes and consequently may be considered by some authors as cases of GC. Moreover, it is generally observed that GC responds poorly to chemotherapy and radiotherapy and has an overall unfavorable prognosis that distinguishes it from LGG. It is thus of the utmost importance to find more objective ways of identifying these GC patients among those presenting with LGG. Subjects and Methods: Single voxel proton MR spectroscopy and 2D-chemical shift imaging using both short (22 ms) and long (135 ms) echo times were performed in 9 patients suffering from GC as compared to 9 subjects with LGG and 25 normal volunteers with the objective to establish the precise metabolic pattern of this uncommon brain neoplasm. In addition, spectra of GC were compared to those of 103 other brain tumors. Results: The gliomatosis infiltration was characterized by markedly elevated creatine and myo-inositol, reduced N-acetyl aspartate and moderately elevated choline. This pattern differs strikingly from LGG which are characterized by elevated choline and myoinositol, markedly reduced N-acetyl aspartate and low to normal creatine. Multivariate MR spectroscopy (MRS) data analysis using principal component analysis clearly separates each tumor groups and enables a clear diagnosis of individual patients. Among our database of more than 100 brain tumors of various histologies no other tumor type exhibits the same metabolic profile as GC. Discussion and Conclusion: While distinction between GC and LGG, based on histology and MRI criteria is a matter of debate, MRS clearly differentiates these two entities and hence the choice of the therapeutic strategy. It also provides new insight into the pathophysiology of GC since elevated creatine and myo-inositol may be related to proliferation of glial elements or more probably activation of normal glia. Elevated choline reflects cellular proliferation and reduced N-acetyl aspartate corresponds to neuronal suffering and/or focal invasion by the tumor process. Moreover, MRS demonstrates that, among all other brain tumors, GC exhibits a specific metabolic profile.

198 Quantification by MRS of glutamate and other brain metabolites in depression before and after repetitive transcranial magnetic stimulation F. Schubert1, F. Seifert1, H. Rinneberg1, M. Bajbouj2; 1Medical Physics and Information Technology, Physikalisch-Technische Bundesanstalt, Berlin, GERMANY, 2Psychiatry and Psychotherapy, Free University Berlin, Berlin, GERMANY.

Introduction: Neurochemical studies suggest that severe depression is accompanied by disorders in L-glutamate (Glu) metabolism [1]. A significant reduction of Glu in the anterior cingulate in depressed patients was detected by proton MRS [2]. As there is growing evidence for a beneficial effect of repetitive transcranial magnetic stimulation (rTMS) in depression, we investigated depressed patients' brains for metabolic changes before and after rTMS, focusing on Glu but also analyzing other metabolites. Subjects and Methods: Eleven patients meeting the DSM-IV criteria for major depression were studied. Antidepressant medication was kept constant 4 weeks prior to and within the treatment time. Exclusion criteria for patients were mental retardation, other significant psychiatric or neurological illness, organic mental disorder, or alcohol or substance abuse within one year before the study. Each patient underwent 10 sessions of 20 Hz rTMS with 1000 stimuli per day over the left dorsolateral prefrontal cortex (dlpfc) with stimulus intensities of 100% of motor threshold. Before and after the rTMS period MR spectra were acquired on a 3Tesla-scanner (MEDSPEC 30/100, Bruker Medical) using PRESS (TR = 3 s, TE = 80 ms; n = 128), from brain voxels including the anterior cingulate (ac, 2.5x4x2 cm3) and the left dlpfc (2x2x2 cm3). Metabolite quantification relies on a recently developed time domain-frequency domain program package [3]. Total Cho (tCho), total Cr (tCr), NAA, Glu, and Gln resonances were fitted, including phantom spectra for the latter three, and imposing prior knowledge of frequency, linewidth and phase. Segmentation of brain voxels was performed using spm99. Results: The figure shows the metabolite levels determined in the two sessions (1, 2). Except for tCho (dlpfc: 1.66 mM, ac: 1.93 mM, p = 0.003) there was no significant difference between the concentrations in the two voxels. The only difference observed between baseline and posttherapeutic concentrations is seen for tCr in dlpfc (p = 0.047), with an increase of the mean concentration from 8.0 to 8.65 mM. Conclusion: The results suggest that rTMS does not induce alterations in the measurable Glu pool in the voxels studied. Furthermore, MRS-detectable metabolism and neuron density appear to remain largely unchanged, thus confirming a recent study on rats[4]. References: 1. Altamura, CA et al, Amer J Psych 150, 1731-1733, 1993 2. Auer, DP et al, Biol Psych 47, 305-313, 2000 3. Schubert, F et al, 10th ISMRM Meeting, Honolulu 2510, 2002 4. Michaelis, T et al, 10th ISMRM Meeting, Honolulu 951, 2002

Brain MRS: Cancer, Psychiatry and Neurodegeneration 199 Localized 1H-spectroscopy at 3 Tesla in depressive patients before and after treatment: a controlled double-blind design S. Gruber1, R. Frey2, A. Stadlbauer1, L. Reinfried2, V. Mlynarik1, S. Kasper2, E. Moser1; 1Institute for Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Department for General Psychiatry, University of Vienna, Vienna, AUSTRIA. Purpose: Treatment of depressive patients with antidepressive drugs might affect metabolic processes in the brain. In order to investigate such effects, a group of previously untreated depressive patients were examined by localized 1H-MRS before and after treatment with Citalopram. The aim of the study was to identify metabolites responding to the treatment. Methods and Materials: 22 depressive patients (36 ± 12 years) were scanned before (=baseline) and after treatment with Citalopram (20-60 mg) and compared with 22 age- and sexmatched healthy subjects in a double-blinded parallel group design on a 3T whole body system (Bruker Medspec S300, Germany) using a standard birdcage coil (STEAM, TE/TR=20/6000ms, 64 averages, 8 cm3 voxels). Absolute concentrations of tCr, NAA, tCho and mI, taking the unsuppressed water signal of the same voxel as a reference, were calculated using LCModel [1] (Version 5.2-2). Group statistics were performed using a t-test for paired samples. Results: In patients and in controls, all spectra showed a signal/noise ratio ≥ 4 and a linewidth FWHM ≤ 0.075 ppm. Compared to controls, depressive patients showed significantly higher absolute concentrations of tCr left-frontally (14%, p=0.02 ) and right-frontally (19%, p=0.008). Four weeks after treatment tCr in patients was lower compared to baseline with a trend leftfrontally ( 8%, p=0.08) and right-frontally (12%, n.s.). However, there were no significant differences between the post-treatment findings and the control group. In the group responding to antidepressive medication tCr decreased in the left fontal lobe (-13%, p=0.058) compared to baseline. Conclusion: For the first time therapeutic effects in depressive patients treated with Citalopram could be monitored using 1H-MRS at 3 Tesla. This may be useful to explore the underlying metabolic mechanism of psychiatric disorders. References: [1] SW Provencher, Magn Reson Med 30:672-679 (1993)

200 Association between serotonin transporter gene polymorphism, hippocampal integrity, and anxiety: An in vivo magnetic resonance spectroscopy study J. Gallinat1, M. Bajbouj2, F. Seifert3, J. Schunter2, C. Wernicke2, H. Rinneberg3, F. Schubert3; 1Psychiatry and Psychotherapy, Humboldt University, Berlin, GERMANY, 2Psychiatry and Psychotherapy, Free University, Berlin, GERMANY, 3Medical Physics and Information Technology, Physikalisch-Technische Bundesanstalt, Berlin, GERMANY. Introduction: The serotonin transporter (5-HTT) gene-linked polymorphic region (5-HTTLPR) has gained interest in anxiety research. Its short (s) allele is associated with lower transcriptional efficiency of the promoter than is the long (l) allele, leading to lower cellular serotonin uptake. On the behavioral level, the s-allele is associated with anxiety-related traits. Serotonin and 5-HTT play a critical role in hippocampal neuronal development and

S107

integrity [1-3]. Furthermore, proper serotonergic signalling during the postnatal period is associated with anxiety-like behaviour in the adult [4]. We investigated the association between the 5-HTTLPR genotype and hippocampal NAA as surrogate marker of hippocampal neuronal integrity, quantified by 1H-MRS, and the role of NAA for anxiety behavior. Subjects and Methods: MRS was performed on the left hippocampus (2x3x2 cm3) and the anterior cingulate cortex (ACC, 2.5x4x2 cm3) in 38 healthy subjects using a 3-Tesla-scanner (MEDSPEC 30/100, Bruker Medical) with PRESS (TR=3s, TE=80ms; n=128). NAA was quantified using a novel time domain-frequency domain program [5]. Anxiety was assessed with the State-Trait-Anxiety-Inventory (STAI). The 5-HTTLPR genotype was determined by standard PCR and gel electrophoresis. Data for subjects carrying s/l and s/s genotype were pooled. Results: ANOVA showed a significant effect of the factor 5-HTTLPR genotype (F=4.08, p=0.021) and the covariate STAI trait (F=5.54, p=0.008) on hippocampal NAA concentration. Subjects carrying the s allele had lower hippocampal NAA concentrations than subjects homozygous for the l allele (11.3mM vs. 12.0mM, p=0.013, F=6.77). No such difference was observed for NAA in the ACC. The figure shows a significant correlation between hippocampal NAA concentration and STAI trait score. Discussion: The 5-HTTLPR genotype appears to be associated with the hippocampal neuronal integrity, compatible with the role of serotonin in hippocampal neurogenesis and cell proliferation [1,4]. The correlation between hippocampal NAA concentration and anxiety scores is in line with the reported link between hippocampal development and anxiety behavior in adult animals [4]. The results suggests that the investigation of intermediate phenotypes (biological markers, e.g. NAA) is an interesting research strategy in medical genetics, since biological markers may represent more proximal gene effects than behavior or diagnostic categories. References: 1. Fabre, V et al, Eur J Neurosci 12, 2299-2310, 2000 2. Mossner, R et al, J Neural Transm,109, 557-565, 2002 3. Yan, W et al, Brain Res Dev Brain Res 98, 185-190, 1997 4. Gross, C et al, Nature 416, 396-400, 2002 5. Schubert, F et al, 10th ISMRM Meeting, Honolulu 2510, 2002

S108

Cardiac Imaging

201 New MRS biomarker for monitoring neurodegenerative diseases: animal models S. Kasparová1, Z. Sumbalová2, P. Bystrický1, V. Mlynárik3, T. Liptaj1, J. Horecký4, J. Kucharská2, A. Gvozdjáková2; 1Central Laboratory of NMR, Slovak University of Technology, Bratislava, SLOVAKIA, 2Pharmacobiochemical lab of 3rd Department of Internal Medicine, Faculty of Medicine, Comenius University, Bratislava, SLOVAKIA, 3Dept. of Radiology, Dérer Faculty Hospital, Bratislava, SLOVAKIA, 4Institute of Preventive and Clinical Medicine, Faculty of Medicine, Comenius University, Bratislava, SLOVAKIA. Introduction: Creatine kinase (CK) plays a central role in energy transfer in cells with high energy requirements and it is very susceptible to oxidative stress. Rate constants of CK reaction can be investigated by magnetization transfer in vivo 31P MRS in rat brains during various modeled chronic form diseases. The rate constant of CK, kfor, was studied in following diseases in adult and aged rat brains: i) alcohol intoxication, ii) Huntington’s disease (HD), iii) diabetes mellitus (DM), and iv) chronic cerebral insufficiency, which modeled Alzheimer’s disease (AD). The aim of these experiments was to demonstrate that the pseudo first-order rate constant kfor of the CK forward reaction is a better indicator of changes in the brain energy metabolism than the conventional steady-state in vivo 31P MRS during chronic pathological states. Subjects: 6 (Adult) and 15-16 (aged) month-old Wistar rats were used in all experiments. Alcohol intoxication was induced by administration of 30% ethanol in adult female rats. HD and DM models were induced in aged rats by administration of 3-nitropropionic acid and streptozotocin, respectively. AD model was prepared by severe hypoperfusion - 3-vascular occlusion (a physiological model) in aged rats. Methods: 31P MRS saturation transfer measurements were performed on a 4.7 T SISCO instrument. The saturation was accomplished by the DANTE pulse sequence. The kfor was calculated according to the equation: kfor = [1 – M*(PCr)/M0(PCr) ] / T1app(PCr), where M0(PCr) and M*(PCr) are the PCr signal intensities in the reference spectrum and in the spectrum after 10-second irradiation of the gamma-ATP resonance, respectively. T1app(PCr) is the apparent longitudinal relaxation time of the PCr signal in the presence of the gamma-ATP saturation. Results: While there were no significant changes in the steadystate of 31P MRS, we found significant changes of kfor in all chronic diseases. Compared to controls (Table 1), kfor decreased in alcohol intoxication and AD, however, this value increased in HD and DM. Discussion and Conclusion: Our findings suggest that the CK reaction could play a key role in energetic system of adult and aged brains in chronic pathological states. We have concluded that the investigation of the kinetic parameters using in vivo MRS magnetization transfer methods can be used as a noninvasive in vivo biomarker for age-related neurodegenerative diseases and can possibly predict energy metabolism damage in brain, which is not yet detectable by conventional MRS methods.

Table 1 Subjects 1.Control, adult rats Alcohol intoxication, adult rats 2.Control, aged rats Huntington's model (HD), aged rats Diabetes melitus (DM), aged rats Alzheimer's model (AD), aged rats

Scientific Session Cardiac Imaging 11:20 am - 1:00 pm

Number of animals n=10 n=7 n=8 n=6 n=6 n=6

kfor (s-1) 0.38+0.02 0.31+0.02 0.30+0.03 0.35+0.01 0.39+0.01 0.20+0.01

T1(PCr)(s) 3.5+0.2 3.2+0.2 3.4+0.3 3.2+0.3 3.2+0.4 3.4+0.1

Ruys

202 Quantification of mitral valve regurgitation with three-directional velocity-encoded MRI J. J. M. Westenberg1, J. Doornbos1, M. I. M. Versteegh2, J. J. Bax3, M. G. Danilouchkine1, H. J. Lamb1, R. J. van der Geest1, A. de Roos1, J. H. C. Reiber1; 1Radiology, Leiden University Medical Center, Leiden, NETHERLANDS, 2Thoracic Surgery, Leiden University Medical Center, Leiden, NETHERLANDS, 3Cardiology, Leiden University Medical Center, Leiden, NETHERLANDS. Purpose/Introduction: Echocardiography, the current tool for diagnosing mitral valve (MV) regurgitation, yields only estimations of the real MV-flow. In this study, a new method is presented for accurate quantification of the MV blood flow, as well as the regurgitant volume, using 3-directional (3-dir) velocity-encoded (VE) MRI. Subjects/Methods: Velocity maps are obtained in a radial stack of 6 slices, positioned along the long-axis of the left ventricle (Figure 1). The velocity vector field of the intra-ventricular blood flow is constructed from the velocity components measured in three directions for 30 phases during the cardiac cycle. The position of the MV-plane is indicated retrospectively on the 2- and 4-chamber views. The MV-flow can be determined from the velocity values reconstructed through this MV-plane. 3-dir VE MRI and 1-dir VE MRI (1) volume flows are compared with the flow through the ascending aorta (AA) measured with MRI. Ten volunteers without valvular disease were included. Reproducibility and accuracy were examined by studying interand intra-observer variations. Also, ten patients with MV-regurgitation, confirmed by transesophageal echo, were analysed. Results: Volume flows measured at the MV with 1-dir VE MRI and 3-dir VE MRI were compared to the volume measured in the AA and presented in Figure 2. 1-dir VE MRI shows an overestimation (p<0.01) of 25 ml on average, compared to the flow through the AA. Also, correlation is very poor (rP = 0.15, p=0.68). 3-dir VE MRI shows no over/underestimation compared to the flow in the AA and there is good correlation (rP = 0.91, p<0.01). One observer performed the analysis of the data twice with an inter-examination time > 1 week and a second observer also performed image analysis. Intra- and inter-observer variations were very small (0.9 ml and 1.3 ml per cardiac cycle) and were not statistically significant. Regurgitation can be quantified and the volumes measured at the MV and correspond with the flow measured at the AA. Discussion/Conclusion: With 1-dir VE MRI, the acquisition plane is positioned only at the MV during end-systole, and therefore, the flow is not measured at the MV during other phases of the cycle.

Cardiac Imaging With 3-dir VE MRI, MV-flow quantification is accurate and reproducible. Inter- and intra-observer variations are not statistically significant. Regurgitation can be quantified with this method. Reference: 1. Kayser HWM. MR velocity mapping of tricuspid flow: correction for through-plane motion. JMRI 1997; 7:669-73.

S109

cle has a more favorable oxygen supply/demand ratio compared to the left ventricle due to its smaller muscle mass. Conclusion: Gd-BOPTA-enhanced MRI in the late phase in patients with acute inferior myocardial infarction can reliably detect RV involvement. At a Gd-BOPTA dose of 0.1 mmol/kg BW a clear delineation of the infarcted right ventricular tissue is possible. Since RV infarction contributes markedly to hemodynamic instability, atrioventricular conduction block and in-hospital mortality, the reliable diagnosis of RV infarction on MRI may contribute to better patient outcome through signaling the need for more intense patient monitoring and observation.

204 Fast Multi-slice cardiac MRI using Controlled Aliasing In Parallel Imaging Results In Higher Acceleration (CAIPIRINHA) M. Blaimer, F. Breuer, M. A. Griswold, R. M. Heidemann, M. Müller, P. M. Jakob; Physikalisches Institut, Universität Würzburg, Würzburg, GERMANY.

203 Gd-BOPTA enhanced MRI in detection of right ventricular involvement in myocardial infarction G. Schneider, I. Kindermann, P. Fries, F. Ahlhelm, B. Kramann, M. Boehm; Dept. of Radiology, Universitaetskliniken des Saarlandes, Homburg/Saar, GERMANY. Purpose: Isolated right ventricular (RV) infarction is relatively rare and accounts for less than 3% of all cases of infarction. However, RV infarction occurs frequently in the setting of an acute inferior myocardial infarction when there is an occlusion of the right coronary artery proximal to the marginal branches. The clinical importance of RV involvement has been shown previously in studies in which the mortality rate after acute inferior myocardial infarction increased from 10% to 47% when an involvement of the right ventricle was present. The aim of our study was to determine whether RV involvement in the setting of inferior myocardial infarction can be reliably detected on late contrast-enhanced images after injection of gadobenate dimeglumine (Gd-BOPTA, MultiHance; Bracco Imaging SpA, Milan, Italy). Material and Methods: 18 patients with acute inferior myocardial infarction and ECG findings indicative of RV involvement were studied by contrast enhanced MR Imaging in the delayed (10 min) phase after Gd-BOPTA injection (0.1 mmol/kg BW) in order to detect late enhancement of the infarcted myocardium. Diagnosis of right ventricular infarction was based on ST segment elevation in lead V4R of the ECG recording. Results: Late enhancement of the inferior aspect of the right ventricle allowing the diagnosis of RV involvement was found in 16 of 18 patients. Six patients had thromboembolic occlusion of the right coronary artery and showed areas of not reperfused tissue in the left ventricle. However, no evidence of not reperfused areas in the right ventricle was seen on MRI. This may have been due to the fact that coronary blood flow to the right ventricle is unique in that it occurs during both systole and diastole, and to the fact that the right ventri-

Introduction: A new approach for fast multi-slice cardiac MRI using partially parallel acquisition (PPA) strategies is presented. The method described here shows improved signal-to-noise behavior and therefore better image quality while keeping the temporal benefits of PPA methods. Theory and Methods: In our first implementation two simultaneously excited slices are shifted by a fraction of the field-of-view (FOV) with respect to each other by applying a different radio frequency (RF) phase cycle for each slice (see Figure 1). In that way the unfolding conditions can be improved significantly which allows an artifact free separation of slices that have basically the same coil sensitivities. Experiments were performed on a 1.5 T Siemens Vision scanner (Siemens, Erlangen, Germany) using a 4-channel receiver body array and a modified segmented FLASH sequence for cine imaging (10 segments, TE / TR = 5 / 10 ms, flip angle = 25°, matrix 140 x 256, FOV 420 mm, 12 frames per heartbeat). Dual band RF pulses were used to excite two adjacent slices (slice thickness 10 mm, distance 20 mm) simultaneously. Image reconstructions were done with conventional GRAPPA [1] using adapted auto-calibration-signal (ACS) lines. Results: In contrast to a conventional multi-slice or 2D-SENSE [2] approach, CAIPIRINHA provides an artifact free slice separation even with directly adjacent slices (see Figure 2). Even though the CAIPIRINHA images are recorded two times faster, they have the same signal-to-noise ratio as images from a standard multi-slice experiment, since the number of phase encoding steps are the same and the geometry factor is close to one at every location. Discussion: The benefits of CAIPIRINHA for cardiac imaging are manifold. As an example, contrast enhanced cardiac perfusion experiments should benefit, because several cardiac slices can be investigated during one first-pass experiment. Furthermore, multi-slice cardiac cine experiments can be performed in shorter scan time without any loss in the signal-to-noise ratio (SNR). In conclusion, CAIPIRINHA offers the possibility to relax the counteractive requirements of imaging speed and SNR met in cardiac MRI. Acknowledgements: This work was funded by the DFG 827/4-1. References: 1. Griswold MA, et al. Magn Reson Med 2002; 47:1202-1210. 2. Weiger M, et al. MAGMA 2002; 14:10-19. 3. Larkman DJ, et al. J Magn Reson Imag 2001; 13:313-317.

Cardiac Imaging

S110

206 High-resolution real-time cardiac SSFP: optimized spatiotemporal sampling for high acceleration with k-t BLAST and k-t SENSE J. Tsao, S. Kozerke, P. Boesiger, K. P. Pruessmann; Institute for Biomedical Engineering, ETH Zurich, Zurich, SWITZERLAND.

205 Normal human right and left ventricular volume, function and mass by cine magnetic resonance imaging with SSFP-puls sequence A. Rauch1, J. Haimerl, MD2, D. Sauer, MD2, W. Bauer, MD1, E. Sauer, MD2; 1Department of Cardiology, University of Wuerzburg, Wuerzburg, GERMANY, 2Department of Internal Medicine, Hospital Landshut-Achdorf, Landshut, GERMANY. Introduction: SSFP- Imaging is a promising technique for cardiac magnetic resonance imaging , as it provides improved blood/myocardial contrast in shorter acquisition times compared with conventional gradient-echo acquisition. Normal values need to be defined for the SSFP-puls sequence because it becomes the preferred cardiac MRI pulse sequence. Subjects and Method: 100 normal subjects (male 50 and 50 female) were examined; weight was 73 ± 13 kg (52 - 105), length was 175 ± 10 cm ( 148 - 201), BSA was 1.87 ± 0.19 m² ( 1.50-2.35) and age was 39 ± 13 y (18-66). A SSFP was used to obtain a stack of parallel short-axis image from the ventricular apex to the base of the heart. The LV and RV volumes and LV mass were calculated by slice summation. Results: Normal ranges were established and indexed to length and body surface area (BSA). Gender- specific differences were found in cardiac volume and mass, but cardiac function showed nonsignificant differences. After indexing for length and BSA gender differences persist for all parameters, except for ESVLV. With age there was a significant decrease in volume and malemass. Function parameters showed an increase. Conclusion: This study provides the base for diagnosis of patholigical findings in patients with heart disease. Table1: LV and RV parameters Female < 40 y (n = 25) ESVLV (ml) 54.8 +/- 16 EDVLV (ml) 118 +/- 21 SVLV (ml) 63.8 +/- 10 EFLV (%) 54.5 +/- 7 MMLV (g) 79.6 +/- 15 ESVRV (ml) 44.8 +/- 14 EDVRV (ml) 103.6 +/- 19 SVRV (ml) 58.9 +/- 11 EFRV (%) 57.5 +/- 9

Female > 40y ( n = 25) 47.2 +/- 12 111.4 +/- 21 64.3 +/- 12 57.8 +/- 6 80.7 +/- 18 37.4 +/- 13 96.9 +/- 22 59.7 +/-12 62.6 +/- 8

Male < 40y ( n= 28) 72.7 +/- 25 158.2 +/- 36 85.6 +/-19 54.9 +/- 7 115.6 +/- 21 64.1 +/- 21 144.3 +/- 29 81.3 +/- 18 56.9 +/- 8

Male > 40y (n = 22) 56.6 +/- 15 135.8 +/- 30 79.2 +/-18 58.5 +/- 5 111.8 +/- 24 59.3 +/- 19 136.6 +/- 30 77.3 +/- 16 57.2 +/- 7

Introduction: Steady-state free-precession (SSFP) has become an indispensable cardiac sequence for its superb blood-to-myocardium contrast. However, real-time SSFP with high spatiotemporal resolutions remains challenging due to its lower sampling efficiency (e.g. vs. echo-planar imaging) and sensitivity to eddy currents (1). In this work, we accelerated SSFP 8-fold using two recently introduced methods, called k-t BLAST and k-t SENSE (2-3) (for single- or multi-coil acquisition respectively), which exploit the inherent spatiotemporal correlations within a time series of cardiac images. To achieve high acceleration, the k-space trajectory was optimized to yield a near-isotropic space-time sampling lattice (4), while gradient discontinuities were minimized to reduce eddy currents. This combination yielded excellent image quality with an in-plane resolution of ~2.0mm at >25 frames/sec. Methods: The theory of k-t BLAST and k-t SENSE is described in ref. (3). Briefly, raw data are acquired by sparsely sampling k-space over time, t. This sparse sampling in “k-t space” leads to a convolution of the object signals in the reciprocal x-f space (x=spatial position, f=temporal frequency). The resultant signal aliasing is resolved by utilizing prior knowledge about the signal covariance of the object. Such signal covariance is learnt from a series of lowresolution training images acquired at a high frame rate. The k-t approach offers new opportunities for optimizing the acquisition. We optimized the k-t sampling pattern by using a lattice (4), which has a sparse point-spread function in x-f space, thus simplifying the reconstruction dramatically. Furthermore, we chose optimized lattices that lead to near-uniform signal tiling in x-f space to minimize signal overlap. Such patterns are feasible only under certain acceleration factors (e.g. 2x,5x,8x,…). Finally, we minimized eddy-current-related problems by traversing the chosen pattern smoothly. Fig.1 shows the optimized lattice for 8x acceleration. Results: Real-time cardiac images were acquired from healthy subjects in a Philips Intera 1.5T scanner (gradient 30mT/m, 150mT/m/ms), using a 5-element phased-array coil. Fig.2 shows SSFP images (a) in short-axis view (voxel size 1.54x2.06x8.00mm3, 27.1 frames/sec) and (b) in 4-chamber view (voxel size 2.00x2.22x8.00mm3, 27.2 frames/sec). Discussion: k-t BLAST and k-t SENSE allow real-time (untrigger, ungated, and free-breathing) cardiac imaging with high spatiotemporal resolutions. By ordering the phase encodes appropriately, the methods can be used with SSFP to achieve high image quality, without eddy-current-related problems. References: 1. Jung BA, et al. MRM 2002;48:921-925. 2. Tsao J, et al. ESMRMB 2002;45. 3. Tsao J, et al. ISMRM 2003;209. 4. Willis NP, Bresler Y. IEEE-TIT 1997;13:208-220.

Cardiac Imaging

S111

208 Noninvasive diagnosis of cardiac shunts in humans using rapid O2-enhanced 1H MR T1 mapping T. Wang1, C. M. Wacker2, A. Rauch2, P. Schanzenbächer2, M. Nahrendorf2, W. R. Bauer2, P. M. Jakob1; 1Department of Physics, University of Würzburg, Würzburg, GERMANY, 2Department of Cardiology, University of Würzburg, Würzburg, GERMANY.

207 Assessment of aortic valve area (AVA), ventricular function and myocardial mass in patients with aortic stenosis (AS) and healthy subjects by magnetic resonance imaging A. Rauch, MD1, J. Dietl, MD2, A. Freitag, MD2, W. Bauer, MD1, E. Sauer, MD2; 1Department of Cardiology, University of Wuerzburg, Wuerzburg, GERMANY, 2Department of Internal Medicine, Hospital Landshut-Achdorf, Landshut, GERMANY. Introduction: Aortic valve area (AVA) is defined as the extent of aortic valve opening in systole. It provides a clinically useful measurement of stenosis severity and can be calculated from Doppler data using the continuity equation or from cardiac catheter (CC) data using the Gorlin equation (GE). Both techniques represent the physiologic orifice area which varies with flow rate. MRI seems to be well suited for quantifying noninvasively the true, anatomic AVA in addition to measures of left ventricular function (LVf), volume and myocardial mass (MM). Subjects / Methods: 15 patients with AS and 25 healthy subjects underwent MRI. The MRI studies were performed on a 1,5 TScanner using a flash 2d and true fisp cine gradient echo sequence. After evaluation of LVf and MM imaging planes were performed to ensure accurate localization of the aortic annulus and valve orifice. The AVA was measured 3 times in 5 mm slices by imaging planes perpendicular to the stenosis jet. The AVA was outlined manually. CC was performed in all patients with AS and the AVA was calculated using the GE. Results: Figure 1,2,3, Table 1 Conclusion: The planimetry of the aortic valve area (AVA) by MRI is a reliable and non-invasively method to quantify aortic stenosis. The MRI results correspond closely to values calculated from catheterization data using the formula developed by Gorlin. Furthermore, MRI visualizes the sequeles of the disease to ventricular function and myocardial mass in one examination. Tab 1: MRI results n

AVA (cm2) EDVLV (ml) ESVLV (ml) EFLV (%) MMLV (g)

Healthy subjects 25 4.0 +/- 0.9 140 +/- 33

67 +/-18

52 +/- 5

Pat. with AS

93 +/-68

22 +/- 20 158 +/- 42

15 1.0 +/- 0.3 151 +/- 61

90 +/- 24

Purpose/Introduction: Quantification of human cardiac shunts is performed usually using noninvasive echocardiography or invasive catheterization by measuring O2 saturations in different heart chambers. Moreover, phase-contrast MRI techniques have been recently established for clinical use. As an alternative, we report a noninvasive approach using a rapid O2-enhanced 1H MR T1-mapping technique (1). Subjects and Methods: Two healthy volunteers and two patients with cardiac shunts, one with left-to-right atrial septal defect (ASD) and another with right-to-left ventricular septal defect (VSD), were examined on 1.5-T VISION, Siemens. Informed consent was obtained before each study. T1 maps in a 4-chamber-view based on an IR Snapshot FLASH technique (1) were acquired in combination with breathhold on end-expiration, while the subjects were breathing 21% O2 and 100% O2, respectively. No ECG triggering was used in all T1-mapping experiments. Molecular O2 physically dissolved in blood is weakly paramagnetic and acts as a T1-shortening contrast agent. Thus, it is possible to obtain information about blood O2 content using molecular O2 as a tracer substance. Results: As an example in Figure 1, (a) anatomical image and (b) corresponding T1 map in a 4-chamber-view from the 55-year-old female ASD-patient while breathing 21% O2 are depicted. ROIs including 30 pixels were drawn in all heart chambers in Figure 1(b) and the corresponding R1 (mean ± SD) was plotted in Figure 2 as a function of inhaled O2 concentrations (2). In this patient, the arterial blood in both left atrium and ventricle show a significant ∆R1 slope of O2 enhancement of 0.16 1/s/%O2 as also observed in healthy volunteers. However, the venous blood in both right atrium and ventricle show a ∆R1 slope of O2 enhancement of approximately 0.06 1/s/%O2 in contrast to near 0 1/s/%O2 in healthy volunteers indicating that the venous blood was not “completely venous” and therefore is mixed with the arterial blood from left chambers via the left-to-right shunt. Discussion/Conclusion: Noninvasive diagnosis of cardiac shunts was successfully demonstrated using a rapid O2-enhanced 1H MR T1-mapping technique. This method directly depicts shunt effects via physically dissolved O2 as a contrast agent. It has the potential to provide qualitative information before and after occlusion of the defect without the need of catheter. Future work will concentrate on quantitative determination of shunt volume. References: 1. Jakob PM, et al. JMRI 2001;14:795-799. 2. Jakob PM, et al. MRM, submitted.

S112

Sequences and Techniques: Imaging and Spectroscopy Results: Simulation: With the higher CM dose and slower injection rate the simulation showed an extended timing window in the myocardial transit-time curve allowing for lower temporal resolution without significant loss of accuracy in perfusion determination. Clinical: According to the 201-Tl SPECT 37 of the total 156 segments were ischemic, 18 infarcted and 101 normal. In the corresponding regions no differences in average perfusion values were found by MRI between normal and ischemic tissue at rest (respectively 0.4 ± 0.21 and 0.4 ± 0.18 ml/g/min, p > 0.9). Significantly lower average perfusion value was found in the infarcted segments (0.27 ± 0.21, p < 0.0001). Discussion/Conclusion: A decrease in perfusion at rest in infarcted regions, according to SPECT scintigraphy, was detected in patients. A slower injection of a higher contrast media dose allows to overcome the spatial limitation of cardiac MR perfusion assessment, and is therefore more advantageous than a short bolus. This protocol presents an important potential for clinical application in MRI evaluation of coronary heart disease.

Scientific Session Sequences and Techniques: Imaging and Spectroscopy 11:20 am - 1:00 pm

Plate

210 209 First-pass cardiac MR perfusion: is slower better? M. K. Ivancevic, MSc1, I. Zimine, MSc1, C. Dornier, PhD1, D. Foxall, PhD2, D. Didier, MD1, A. G. Somsen, MD, PhD3, A. Righetti, MD3, J. Vallée, MD, PhD1; 1Radiology, Geneva University Hospital, Geneva, SWITZERLAND, 2MR Clinical Science, Philips Medical Systems, Cleveland, OH, 3Cardiology, Geneva University Hospital, Geneva, SWITZERLAND. Purpose/Introduction: In patients with coronary artery disease assessment of myocardial perfusion is of clinical and prognostic value. MR first pass perfusion imaging has become a widely used clinical tool for assessing myocardial perfusion. This study is a clinical validation of a new protocol consisting in a slower contrast media (CM) injection and image sampling of a higher CM dose, in order to overcome the spatial coverage limitation. The efficiency of the protocol was demonstrated in patients with a history of myocardial infarction using 201-Tl SPECT as reference. Subjects and Methods: Twelve patients (all male, mean age 60±11 years) with stable coronary artery disease and no acute symptoms underwent a rest/effort 201-Tl SPECT and a rest MR exam. MR imaging was performed on an Eclipse 1.5T MR system (Philips Medical Systems, Cleveland, OH) with an RF-FAST sequence and following parameters: TI/TR/TE 28/3.74/1.5 ms, 50 kHz bandwidth, 40° FA, 90°-180° preparation pulses, and a 112x128 matrix. A cardiac surface coil and ECG trigger were used. Eight slices were acquired during three to six heart beats, depending on patients heart rate. A bolus of 0.08 mmol/kg Gd-DTPA was injected in a brachial vein at 0.5 ml/s rate followed by 10 ml of isotonic saline with a Medrad Spectris MR injector. The myocardial response and the effect of the bolus on the perfusion values were evaluated in a simulation using the one-compartment model.

Fast 3D SSFP based proton SI with oscillating readout gradient M. Althaus1,2, W. Dreher1,2, C. Geppert1,2, D. Leibfritz1,2; 1FB2 (Chemistry), University of Bremen, Bremen, GERMANY, 2Center for Advanced Imaging, University of Bremen, Bremen, GERMANY. Introduction: A fast 3D proton spectroscopic imaging (SI) pulse sequence based on steady state free precession (SSFP) methods is presented. Different to the recently proposed spectroscopic variant [1] of CE-FAST [2] the signal is sampled in presence of an oscillating read gradient. Thus, the already short minimum total measurement time Tmin can be further reduced. This sequence is of interest especially at high B0 because of the short Tmin, the high SNR and the increased spectral resolution. Subjects and Methods: The modified spectroscopic CE-FAST pulse sequence is shown in Figure 1. Following a chemical shift selective composite pulse the echo-like signal is acquired in presence of an asymmetric oscillating read gradient [3-5]. The sequence was implemented on a Bruker Biospec 4.7T/40cm system. Measurements were performed on a spherical phantom filled with an aqueous solution of N-acetyl aspartate (NAA, 25mM) and myo-inositol (Ins, 75mM) and healthy male Wistar rats. The measurement parameters were as follows: voxel size 1.5x3x4mm3 (phantom)/1.5x3x3mm3 (in vivo); matrix (64, 64,32,16) in (kx,kω,ky,kz); flip angle 40°; TR=76ms; SW=1600Hz achieved by two interleaved acquisitions. The total measurement time was 3:28min. SI spectra are visualized as phase corrected spectra. Metabolic images are calculated using peak area integration. Results: Figure 2a shows an Ins map of the central slice of the spherical phantom. The signal is distributed homogeneously within the phantom (air bubble at the top). Figure 2b shows a representa-

Sequences and Techniques: Imaging and Spectroscopy

S113

tive phase corrected spectrum from a central voxel with both the singlet signal of NAA (2.02ppm) and signals from the J-coupled spins of Ins (3.28ppm, 3.54ppm, 3.60ppm, 4.05ppm). A FLASH image and an array of NAA spectra from a central slice of the rat brain in vivo are shown in Figure 3a. In the phase corrected spectrum of one voxel (Figure 3b) the signals are assigned to various metabolites (NAA, Ins, tCr, Cho+Ins+Tau). Discussion: A fast 3D SSFP based SI sequence with an EPI readout module is presented. As shown by phantom and in vivo measurements, signals from uncoupled and J-coupled spins can be measured within a few minutes. The short Tmin and the large matrix size makes this technique appealing to applications where both 3D spatial and high temporal resolution are needed. References: 1. Dreher W et al [in press] MRM 2. Gyngell ML [1988] MRI 6:415-419 3. Mansfield P [1984] MRM 1:370-386 4. Hennel F et al [1995] MRM 34:520-524 5. Posse S et al [1994] Radiology 192:733-738

211 Paediatric headcoil for optimal 1H MR spectroscopy at 3 Tesla D. W. J. Klomp1, M. van der Graaf1, M. A. A. P. Willemsen2, Y. M. van der Meulen1, A. P. M. Kentgens3, A. Heerschap1; 1Radiology, University Medical Center Nijmegen, Nijmegen, NETHERLANDS, 2Paediatric Neurology, University Medical Center Nijmegen, Nijmegen, NETHERLANDS, 3Physical Chemistry, University of Nijmegen, Nijmegen, NETHERLANDS. Introduction: Metabolic information from paediatric patients can be acquired by 1H MR Spectroscopy. To avoid partial volume effects, like mixture of grey and white matter, a small region of interest should be selected. Therefore the signal to noise ratio (SNR) has to be optimised. The SNR can be increased by going to higher field strengths. However, this leads to higher spectral bandwidths, which increases the chemical shift artefact, and also to a lower T2. Although STEAM localisation can be used with rather short echo-times (TE) and reduced chemical shift artefacts, it intrinsically leads to a 50% lower SNR compared to PRESS localisation. We therefore used PRESS localisation with high RF bandwidths that minimises the chemical shift artefact. Since these pulses are smaller, short TE can be maintained as well. A dedicated paediatric headcoil has been designed for 3T that can handle the large peak power required for the high bandwidth RF pulses in the PRESS sequence. Methods: An 8 leg circularly polarised coil was designed with an inner diameter of 19cm. The coil can be manually tuned and matched for different loads (i.e. children from 0 until 8 years old). The coil is interfaced to a clinical 3T MR scanner (Trio, Siemens, Erlangen) using a home-built transmit and receiver switch with integrated preamplifier. The efficiency of the coil was compared with the original Siemens headcoil, using the value of the reference power on the same patient. The standard PRESS sequence was modified so that the bandwidths of the RF pulses were scaled until the full available peak power was used (i.e. 8kW for local coils). PRESS (TE=20ms) brain spectra were acquired from small (3.4ml) regions of several paediatric patients.

S114

Sequences and Techniques: Imaging and Spectroscopy conditions, e.g. due to fatigue during studies of exercising muscle. An interleaved STEAM sequence was developed to acquire localised 1H and 31P spectra from two arbitrarily positioned voxels in one and the same experiment. Subjects and Methods: A pulse sequence for interleaved acquisition of 1H- and 31P (or any other nucleus) spectra with STEAM localisation was implemented on a 3T Bruker Medspec whole-body scanner on the basis of MultiScanControl-Tool. Size, shape and position of the two voxels (see Fig.1) and other sequence parameters can be chosen independently.

Results: The efficiency of the coil increased more than twofold compared to the standard Siemens headcoil based on the reference power. This was confirmed by checking the SNR of the images (figure 1). At this 4-year old patient, the B1 field strength could be increased to 86µT using the full 8kW RF power. This led to an increased bandwidth of 4kHz using 7 lobe MAO refocusing pulses. Figure 2 shows a PRESS (TE=20ms) spectrum of white matter of the same patient, which shows good SNR and spectral resolution. Conclusion: A high power Transmit/Receive coil was used to acquire short echo-time PRESS localised 1H spectra of small regions in the brain at 3T with increased RF bandwidths.

The number of excitations/acquisitions of 1H spectra per 31P excitation can be adjusted arbitrarily in integer steps to optimise TR for the respective metabolites. T1 of 1H metabolites is typically on the order of 1s [1,2], while 31P metabolites have significantly longer T1s, i.e. 3-6s at 3T [3]. Therefore, a reasonable acquisition ratio is 1H:31P=4:1 (yielding TRs of e.g. 2s:8s). CHESS scheme was used for water suppression. Experiments on a volunteer's calf muscle were performed using a double tuned 1H/31P surface coil (Ø=10cm) for transmit and receive. Results: The performance of the interleaved 1H/31P STEAM was successfully verified using a two-compartment test object [3]. In vivo spectra (localisation shown in Fig.1) acquired with standard STEAM experiments (top) and with the interleaved 1H/31P STEAM (bottom) sequence are equivalent (see Fig.2).

212 Interleaved acquisition of 1H/31P spectra with STEAM localisation at 3 Tesla M. Meyerspeer1, M. Krssák2, E. Moser1; 1Institute for Medical Physics, Vienna University, Vienna, AUSTRIA, 2Internal Medicine 3, Vienna University, Vienna, AUSTRIA. Purpose/Introduction: Magnetic resonance spectra of different nuclei can provide complementary information on tissue metabolism. Acquiring two sets of spectra in consecutive experiments is both time consuming and may result in undesirably modified test

Repetition time was TR=2s for 1H spectra (left) and TR=8s for 31P spectra (right). Other acquisition parameters: TE=8ms, TM=30ms, BW=2500Hz, 1024 data points (zero-filled to 8k), 16 averages. The 1H VOI was 4cm2 and the 31P VOI was 45cm2 to partly compensate for the lower sensitivity. Discussion/Conclusion: Interleaved acquisition of STEAM spectra of 1H and 31P is shown to yield results equivalent to the standard single nucleus acquisition on a test object and in vivo.

Sequences and Techniques: Imaging and Spectroscopy The sequence allows a number of proton spectra to be acquired during the (necessarily long) 31P TR. Future developments may be the implementation of multiple TEs for protons during one 31P TR, lactate editing or adiabatic water suppression. References: 1. Krssák M, Mlynárik V, Meyerspeer M, Moser E, Roden M [2003] ESMRMB (submitted). 2. Mlynárik V, Gruber S, Moser E [2001] NMR Biomed 14:325-331. 3. Meyerspeer M, Krssák M, Moser E [2003] Magn.Res.Med. 49:620-625.

213 Diastolic tagging indicates effectiveness of endurance training to left ventricular relaxation S. M. Kivistö1, M. A. Perhonen2, K. Lauerma1; 1Radiology Clinics, Helsinki University Hospital, Helsinki, FINLAND, 2Cardiology Clinics, Helsinki University Hospital, Helsinki, FINLAND. Purpose: To assess the effect of endurance training to left ventricular (LV) diastolic and systolic myocardial function with fourchamber-view tagging magnetic resonance imaging (MRI). Subjects and Methods: Long QT syndrome (LQTS) causes failure in ventricular relaxation and predisposes to cardiac arrhythmia (Vincent M [2000] Cardio Clin 18:309-325). Six genetically confirmed, asymptomatic LQTS patients (mean age 40, range 23-54, two male and four females) and eight healthy controls (mean age 45, range 29-58, two males and six females) were examined with tagging and cine MRI (1.5T Siemens Sonata, Erlangen, Germany) before and after three-month endurance training. Four-chamber-view for tagging was set perpendicular to twochamber-view gradient echo images. The four-chamber view gradient echo images (repetition time=51msec,echo time=1.5msec, flip angle=20, matrix=256x256, field of view=320mm) were labelled with cross-section lines at 90 degrees angle to the interventricular septum (Figure 1). The distance between taglines was 15mm. Diastolic and systolic tagged cine series were imaged. LV short axis cine series (repetition time=34msec, echo time=1.5msec, flip angle=60, matrix=256x256, field of view=360) were obtained for the assessment of volumetric data of LV at every 15mm from LV basis to apex. Images were manually traced with NIH image program (http://rsb.info.nih.gov/nih-image). The distance between basal and apical lines (=distance between six lines) on LV septum and posterior free wall was determined from tagged images. The distance between three apical and three basal lines were determined separately to assess regional movement in LV septum and posterior free wall. Longitudinal diastolic relaxation and systolic contraction (mm) and velocities (mm/sec) were determined. Results: Asymptomatic LQTS patients had similar LV diastolic relaxation as controls before training period. Endurance training improved LV relaxation and systolic function similarly in both groups. Diastolic longitudinal relaxation improved significantly in LV septum and posterior free wall (Table 1). Regional analysis of LV septum indicated improvement in diastolic relaxation in basis (P=0.005) and systolic contraction in apex (P=0.007). Heart rate reduced, LV mass increased, end systolic volume reduced and ejection fraction increased (Table 2). Conclusion: Myocardial diseases alter LV diastolic function before any changes can be observed in systolic function (Takemoto K,

S115

Bernstein L, Lopez J, et. al. [1992] Am Heart J 124: 143-148). Evidently training improves diastolic function before the effect on systolic function of myocardium. It is important to develop new methods to study LV diastolic relaxation. The four-chamber-view tagging MRI method is a promising clinical application to estimate effects of training on LV myocardial diastolic relaxation. Table 1: Four-chamber-view tagging MRI results in LQTS patients (n=5) and healthy controls (n=8) Septum Left ventricular free wall Before After PBefore After Ptraining training value training training value Diastolic relaxation mm 7.7+2.9 11.5+4.2 0.0005 7.9+2.6 9.8+3.0 0.0102 mm/sec 20.9+9.7 30.3+12.5 0.0013 21.0+7.3 25.9+9.4 0.0305 Systolic relaxation mm -7.5+3.2 -9.2+2.5 0.0475 -8.4+4.7 -9.7+3.0 0.1225 mm/sec -29.2+15.6 -35.3+17.8 0.141 -31.7+18.0 -38.3+15.6 0.105 Table 2: Left ventricular volumetric data in LQTS patients (n=6) and healthy controls (n=8) Left ventricle Before training After training P-value Heart rate /min 71+7 60+8 0.004 End diastolic volume ml 121+19 125+23 0.141 End systolic volume ml 47+11 42+13 0.008 Ejection fraction % 62+5 66+7 0.004 Mass g 97+19 105+18 0.0008

214 MR-imaging of biocompatible materials: Resolution and delineation of solidification A. Berg1, W. Kluger2, E. Moser1,3; 1Department of Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Department of Conservative Dentistry, Dental School, University of Vienna, Vienna, AUSTRIA, 3Department of Diagnostic Radiology, University and General Hospital of Vienna, Vienna, AUSTRIA. Introduction/Purpose: MRI is known for offering high spatial resolution and contrast in soft tissue. Solid type materials in the human body as human dental tissue or polymer type implants give no active MR-signal due to their very short T2 relaxation time T2. Thus polymer implants can only be visualized as hypointense foreground versus a bright background. Their inner structure including solidity reduced regions cannot be investigated. Single-point-imaging (SPI) enables imaging of semi-solid materials in the human body1. In this contribution we perform SPI on composites based on polymers and ceramic compounds as used in conservative dentistry. We investigate whether it is possible to delineate the polymerization process accompanying illumination of the composite paste and visualize incomplete solidification.

S116

Sequences and Techniques: Imaging and Spectroscopy

Materials and Methods: High resolution along with short detection-time TD in SPI is achieved at 3T (BRUKER MEDSPEC S300) using a powerful 20cm-gradient-system (G=200mT/m) in combination with a 1H-free 15mm birdcage-resonator (voxel-size: 364x364x438µm3). Quantification of the polymerization process is obtained via T2*, which is obtained from a uniexponential fit to the SPI-image with different detection-times (TD=140,200,300µs). Errors are obtained from regression-analysis. The composite is composed mainly of BIS-GMA and a mixture of barium-aluminum-fluoride with silicon-dioxide as filler and used for conservative dentistry ("Charisma Syringe", Heraeus-Kulzer, Dormagen, Germany). Small (12x12x2mm3) disks are illuminated with different exposure-time between 5s and 20s. Results: An increase of the relaxation-rate R2*=1/T2* with exposure time is observed (fig.1) indicating the polymerization process in the composites. For investigating spatial resolution we irradiated the composite paste via an absorption mask consisting of 270µm and 520µm periodic openings in a brass-grid. The periodic modulation in the T2*-weighted SPI-image due to incomplete solidification under the absorption finger can be observed, enabling qualitative access to the resolution obtained (fig.2).

Figure 1. SPI-obtained relaxation rate R2* in composite disks with different light exposure for solidification.

Figure 2. T2*-weighted SPI image of an irradiated composite disk. Dark stripes indicate the solidification process under the brass grid (half period a=500µm) to estimate the resolution obtainable. Conclusion: The solidification process in polymers can be delineated by T2*. The completely hardened composites however cannot be visualized. A resolution below 600µm in incomplete

solidification might be obtained but is actually depending on the individual T2* of the biocompatible material and gradient-strength available. References: 1. A. Berg, N. Dragostinoff, M. Aoid, W. Kluger, E. Moser; EMBEC 2002 Vienna/Austria; IFMBE Proceedings Series Vol. 3/2002 (ISSN 1680-0737)

215 Ultra-high temporal resolution cardiac function in a single breath-hold D. A. Herzka, J. A. Derbyshire, P. Kellman, E. R. McVeigh; Laboratory of Cardiac Energetics, National Heart Lung and Blood Institute, Bethesda, MD. Introduction: A clinical cardiac magnetic resonance (CMR) exam used to evaluate cardiac function acquires a series of images of a slice of the heart at different times during the cardiac cycle. Previously, the temporal resolution of these cine scans was ~3050ms, limited only by the breath-hold length typically 5-20s. This temporal resolution is sufficient to study general cardiac function but it precludes the detailed study of extremely rapid mechanical cardiac phenomena. To optimally capture tissue motion the temporal and spatial resolutions should be matched to the maximum expected velocity. Based on a maximum myocardial velocity of 10cm/s during diastolic filling, tissue can move one pixel (~1.4mm) in 14ms, suggesting that the current temporal resolutions >30ms are insufficient to fully assess the motion. We propose a technique for the acquisition of a cine loop of the motion of the heart with temporal resolutions <5.5ms in a single, reasonable breath-hold. Methods: Imaging was performed in a 1.5T GE scanner using an 8-channel phased array coil and a custom built 8-channel receiver. To achieve maximum acquisition efficiency a 3-echo steady-state free precession (SSFP) imaging sequence with hardware optimized gradient waveforms was used for all acquisitions. Matrix sizes of 256x144 and 192x144 with 36x27cm FOVs were acquired, yielding in plane resolutions of 1.4x1.9mm and 1.9x1.9mm, with slice thickness of 8mm and with temporal resolutions (TR) of ~5.4 and ~4.5 ms respectively. A parallel imaging technique (TSENSE) was used to accelerate the acquisition by a factor of 2-4 with an expected decrease in SNR. Upon detection of an ECG trigger imaging was performed for 90% of the cardiac cycle under breath-holding. A fat saturation pulse applied at the end of imaging and a linearly increasing flip angle series were used to alleviate artifacts at the initiation of imaging. The first imaging heartbeat was discarded. Results: Cardiac cine loops with whole heart coverage including 2 and 4-chamber views were acquired in 3 normal human volunteers. Image quality was excellent and ghosting from the use of TSENSE was minimal. The cine loops revealed motion not typically visible in regular cine imaging: e.g., vibrations from closing and opening of valves. Conclusion: A CMR cine imaging technique with a temporal resolution of ~5 ms is presented. The breath-hold lengths were as short as 13 heartbeats, making the acquisitions possible on patients.

Sequences and Techniques: Imaging and Spectroscopy

S117

216 Contrast behavior of conventional and hyperecho-TSE's at 1.5T and 3T: comparing relaxation effects M. Weigel, J. Hennig; Section of Medical Physics, Department of Diagnostic Radiology, University Hospital of Freiburg, Freiburg, GERMANY. Purpose: Multi-spinecho-sequences such as TSE are widely used in diagnostic MRI, despite their high RF power deposition. Recently, new RF pulse schemes like hyperechoes[1] and TRAPS[2] can be implemented in TSE-sequences (hyperTSE-sequences) to reduce SAR while maintaining SNR. However, since stimulated echo pathways contribute in all lower flip angle sequences, contrast changes subtly. This contribution presents first quantitative studies on the different relaxation behavior of conventional and hyperTSE-sequences to provide a more detailed analysis of such techniques. Subjects and Methods: All experiments were performed on two whole-body imaging systems (1.5T/3.0T, Siemens Sonata/Trio, Siemens, Erlangen, Germany). Different schemes of flip angle variations were implemented in a common TSE-sequence (MTX=256x208, FOV=(30cm)2, ESP=6.6ms, ETL=27, slth=5mm). Besides the conventional modes with constant flip angles (180°, 90°, 60°), the mechanisms “hyperechoes” and “TRAPS” (both with flip angles 90° and 60°) were implemented. Simple linear and optimized sinusoidal flip angle variations were used for TRAPS. Relaxation behavior was quantified for different echo times TE (73ms, 86ms, 100ms, and 113ms) and signal intensities were evaluated in ROI's of cerebral WM, GM, CSF, and orbital fat. Results: Fig.1 shows an exemplary study (TE=73ms) of the relaxation behavior. TSE-sequences with low constant flip angles show reduced signal intensity and contrast, which is even more pronounced at 3T (Fig.2). Both, hyperechoes and TRAPS, reinstate full intensity or even more in all tissue-types. The contrast between WM and GM is reduced compared to TSE180° for conventional low flip angle TSE's and to a lesser degree for hyperechoes. This effect is more pronounced at 3T. Discussion and Conclusion: The signal intensities for WM and GM are higher with the hyperecho- or TRAPS-principle than with the TSE180°. They are even higher if low refocusing flip angles are used in connection with hyperechoes or TRAPS. This is due to T1>>T2, which is particularly valid for high fields. The contrast is reduced for constant low flip angle TSE's and for the hyperecho data, especially at 3T, which should be compensated for using longer effective echo times. In conclusion, hyperecho and TRAPS with identical nominal echo time images yield SNR comparable to TSE180°, similar contrast for the TRAPS principle and reduced effective echo time for hyperechoes. However, RF power deposition can be reduced by as much as 75% by using hyperTSE-sequences [1,2]. 1. Hennig J, Scheffler K, MRM 46:6-12 2. Hennig J, Weigel M, Scheffler K, MRM 49:527-35

217 Comparison of three-point Dixon Chemical-Shift Sequence with Quad-Field-Echo Sequence on a Low-Field-Strength Open Magnet for evaluating articular cartilage defects P. P. Arcuri, V. M. Arcuri, F. Serrao; Radiology, Pugliese-Ciaccio Hospital, Catanzaro, ITALY. Purpose/Introduction: Few studies have specifically addressed the capabilities of low-field-strength MR imaging in the evaluation of articular cartilage defects (1,3). A method for fat suppression using a low-field magnet is the phase-contrast method, described by Dixon (2) that is based on the chemical-shift phenomenon. An alternative imaging method using a low-field magnet is the modified three-point Dixon sequence with a single radiofrequency echo single-scan for water and fat separation (4). The purpose of our study was to compare the diagnostic ability of the three-point Dixon sequence with Quad-Field-Echo sequence for evaluating articular cartilage defects on a 0,35 T open magnet, correlated with artroscopy.

S118

Sequences and Techniques: Imaging and Spectroscopy

Subject and Methods: We prospectively studied 167 patients. The images were obtained with an 0,35 T open magnet (OPART, Toshiba). Were studied patellar and talus cartilage. Imaging parameters were: Quad Field-Echo 2D (TR/TE, 834/40; flip angle, 23°) and Water-Fat Spin-Echo Three-Point Dixon (TR/TE, 1200/36). Results: The axial plane showed higher sensitivity and specificity (respectively 77% and 89%) than sagittal plane (63% and 71%). Were identified 64 cartilage defects: 18 grade I, 37 grade II, 7 grade III and only 2 grade IV. Three-point Dixon sequence showed higher sensibility (81%) than Quad-Field-echo sequence (72%) in detection grade I and II of cartilage defects (fig. 1, 2). In lesions with involved sub-condral bone (fig 3 ) and lesion grade III (fig 4) Quad-Field-Echo recorded a sensibility of 78% (as against Dixon 89%). The sensibility in detection IV grade lesions was substantial the same (89% Quad and 91% Dixon). Discussion/Conclusion: At a low field strength, the time needed for the water and fat signals to transition from in-phase to out-ofphase is longer than at a high field strength. Therefore, both inphase and out-of-phase data can be acquired after a single excitation, in contrast to the other Dixon methods that required 2 or 3 excitations. Our Dixon method uses a single radiofrequency echo, which leads to a considerable reduction in imaging time for this sequence. Our study showed that cartilage defects can be evaluated reliably with low-field-strength MR imaging using the modified three-point Dixon sequence. References: 1. Ahn JM, Kwak SM, Kang HS [1998] Radiology. 208:57-62. 2. Dixon WT [1984] Radiology. 153:189-194 3. Scoot B, Norbert J, Marcus T, Garry E [2003]. AJR. 180:357362. 4. Zhang W, Goldhaber DM, Kramer DM [1996] J Magn Reson Imaging. 6: 909-917.

Epilepsy

S119

and AMY isotropic diffusion map, ADC values were averaged from four rostro-caudal DWI partitions using ROIs of constant size (20mm2). ROIs were drawn according to each subject’s anatomical landmarks as observed from the original T2-DWI image and the 3D-T1data set. Mean ADC values were compared in-between groups and correlated with measures of quantitative MRI. Results: Mean ADC values for the populations studied are depicted in Table.

Table: Mean ± 1SD for controls and patients with right and left seizure focus. Correlations with ipsilateral and contralateral normalized volumes and relaxometry values are shown on charts 1-4.

Clinical Focus Session Epilepsy 3:00 pm - 4:00 pm

Willem Burger

218 ADC mapping of temporal lobe epilepsy: focusing on the hippocampus and the amygdala P. M. Goncalves Pereira1, M. Forjaz Secca2, E. Oliveira3, C. Jordão4, P. Evangelista5; 1S. Neurorradiologia, H. Egas Moniz, Lisboa, PORTUGAL, 2Physics Department, Fac. de Ciencias e Tecnologia / UNL, Lisboa, PORTUGAL, 3Dep. Estatistica, ISCS, Lisboa, PORTUGAL, 4Rm-sdi, Ressonância Magnética de Caselas, Lisboa, PORTUGAL, 5Rm - sdi, Ressonância Magnética de Caselas, Lisboa, PORTUGAL. Introduction: Mesio-temporal damage is a common feature in temporal lobe epilepsy (TLE). Areas such as the hippocampus (HC) and the amygdala complex (AMY) often display a pattern of micro-structural disorganization with selective neuronal loss and gliosis. Diffusion-weighted imaging (DWI) is particularly suited to depict brain regions where the diffusion of water molecules is impaired (quantified by its ADC). The purpose of our investigation was to study the interictal ADC characteristics in the HC and AMY of epileptic patients as compared to normative references and correlate obtained with volumetry and T2-relaxometry. Methods: 30 chronic adult TLE patients and 20 age-matched controls were included. DWI was obtained at b=1000 on a coronal plane perpendicular to the main TL axis. Additionally, quantitative MR-based coronal T1-volumetry (HCVOL) and T2-relaxometry (HCT2) were performed [1, 2]. To obtain each individual mean HC

Chart 1-2: Correlation with HC VOL (r=-0.71; ipsilateral) and HC T2 (r=0.79; ipsilateral). Doted lines mark the control mean ± 1SD. Open circles are contralateral values.

S120

Epilepsy 219 An fMRI protocol to evaluate both language and memory for pre-surgical evaluation in temporal lobe epilepsy N. Danchaivijitr1, A. Zaman2, J. Martinovic3, A. Mayes3, C. Danchaivijitr4, U. Wieshmann5, N. Roberts2; 1Radiology, The Walton Centre for Neurology and Neurosurgery (WCNN), Liverpool, UNITED KINGDOM, 2Magnetic Resonance and Image Analysis Research Centre (MARIARC), University of Liverpool, Liverpool, UNITED KINGDOM, 3Department of Psychology, University of Liverpool, Liverpool, UNITED KINGDOM, 4Neurosurgery, The Walton Centre for Neurology and Neurosurgery, Liverpool, UNITED KINGDOM, 5Neuroradiology, The Walton Centre for Neurology and Neurosurgery (WCNN), Liverpool, UNITED KINGDOM.

Chart 3-4: Correlation with AMY VOL (r=-0.18; ipsilateral) and AMY T2 (r=0.22; ipsilateral). Discussion: Epileptic hippocampal and amygdaloid damage are expressed by changes in DWI. Increased ADC values correlate with the progressive disorganization of the structure depicted by quantitative-MR. This observation was prominent in the sclerotic hippocampi. ADC analysis can be used has a sensitive tool to confirm seizure lateralization and access the degree of hippocampal and amygdala damage. References: 1. Gonçalves Pereira, PM, et al. [2002] Proceedings 10th ISMRM: 202. 2. Gonçalves Pereira, PM, et al. [2002] Proceedings 10th ISMRM: 201. Support: Fundação para a Ciência e a Tecnologia BD 18498/98 and Fundação Grünenthal

Purpose: Language and memory lateralisation assessment is an essential component of an epilepsy surgery programme. A standard, but unfortunately invasive, investigation know as the Wada test is often necessary. The objective of this project is to develop a comprehensive, quantitative and reproducible fMRI protocol to evaluate language and memory lateralisation in patients with temporal lobe epilepsy (TLE) and to compare with the results for the Wada test. The language paradigms are designed to identify both expressive and receptive regions. The memory assessment are created to respectively preferentially activate the medial temporal lobes. Materials and Methods: fMRI was performed using a 1.5T GE (Milwaukee, USA) LX System. BOLD contrast T2*-weighted EPIs were acquired encompassing the whole cerebrum. To enable anatomical localisation of functional activation, a T1-weighted 3D inversion recovery prepared gradient echo image was also acquired. Six two conditions box-car paradigms were employed: (i) Verbal Fluency task (generation and articulation), (ii) Rhyming task (phonological processing and articulation), (iii)Sentence Comprehension task (true/false judgment), (iv) Complex Scenes task (indoor/outdoor judgment), (v) Pattern Memory task (spatial memory encoding), (vi) Word-Association and memory task (word encoding). The order of the paradigms was counterbalanced between subjects. Relevant behavioural data were acquired for paradigms (ii), (iii), (iv), (v) and (vi) during both active and baseline conditions. Postscan memory tests were performed for (iv), (v) and (vi). Data were analysed using SPM99 software. In particular, single subject (p<0.001) and group conjunction (p<0.05) analysis were performed. Results: Analysis of the behavioural data revealed that all the subjects were able to perform the tasks successfully. Preliminary group data analysis revealed that all three of the language paradigms resulted in activation of inferior frontal gyrus (IFG) (Broca's). Both the Rhyming and Sentence Comprehension tasks additionally produced significant activation of left superior temporal gyrus (STG) (Wernicke's). Interestingly, the Sentence Comprehension task also activated right IFG. Both the Complex Scenes and Pattern Memory tasks produced bilateral parahippocampal activation whereas for the Word Association Memory task this was left-sided. Conclusion: Our results demonstrate that both the Rhyming and Sentence Comprehension tasks, in addition to the Complex Scene and Word Association memory tasks will be most suitable to incorporate into a clinical fMRI protocol to implement with TLE patients.

Epilepsy We are currently applying this protocol in patients with temporal lobe epilepsy for whom we have already obtained detailed measurements of brain structure, including hippocampal volumes and T2 relaxation time data.

220 Verbal processing in patients with left hemisphere lesion: A fMRI study J. Tintera1, J. Klener2, P. Krsek3, R. Chabiniok1, J. Vymazal4; 1Department of Radiology, IKEM, Prague, CZECH REPUBLIC, 2Department of Neurosurgery, Hospital Na Homolce, Prague, CZECH REPUBLIC, 3Department of Child Neurology, Charles University, Motol faculty Hospital, Prague, CZECH REPUBLIC, 4Department of Radiology, Hospital Na Homolce, Prague, CZECH REPUBLIC. Introduction: Functional MRI has been shown as very useful to study verbal processing in the presence of various left hemisphere pathologies. The index of laterality is usually used to describe the role of left hemisphere dominancy (defined as the difference between number of active voxels in left and right hemisphere divided by the total sum of all active voxels). This approach is ideal to measure the global shift of the laterality but it cannot describe local changes due to the pathology influence. Therefore, unlike in previous papers we decided to observe changes in verbal activity according to activation clusters of normal verbal fluency. In patients we describe shift or replacement of activity in respect to these regions qualitatively. Method: fMRI measurements were done on Siemens Vision 1.5T with GE EPI sequence (TE=54 ms, TR=4 s, α=90°). Block design was used: 27 slices (THK=4 mm), 64 images, 8 images per each period of rest and stimulation. Subjects performed silent counting during rest and word generation during stimulations. Evaluation was done in SPM 99: Realignment, smoothing and normalization, square-box function (uncorrected p=0.001). A group of 10 healthy right-hander volunteers was used to create standard map of verbal activity. Second order group statistic mapped the normal activation and 3 strong clusters were found (Tailarach coordinates): L1=[-51,27,11] (Brocca), L2=[-50,26,24], L3=[-53,13,35], symmetrically R1, R2, R3 (fig. 1). Patients were assessed in the context of the presence, shift or replacement of activity in respect to these regions. Two groups of patients with left hemisphere pathology were studied: A) 12 young patients - age 9-21 years, B) 8 older patients - age 29-71 years.

S121

Results: Table 1 and 2 summarize results from group A and B respectively. There is higher probability to replace verbal processing regions into the right hemisphere in young patients (group A), usually regions close to the pathology are mirrored or replaced (patients 1,2,6,10). In group B only 1 patient (6) demonstrated high symmetry of the activation. However this patient suffered from congenital blindness and also likely primary lesion could have strong impact on the functional organization during long time period.

Discussion: We confirm that brain verbal functionality is easily reorganized in young patients in presence of left hemisphere pathology. Moreover, in some cases only these regions having close relation with the lesion are replaced into the right hemisphere while tissue with normal function can stay uninfluenced. Study was supported: MZCR IGA NF-6377-3/2000

221 Pre-surgical evaluation of epilepsy by means of qualitative and quantitative MRI P. M. Gonçalves Pereira1, C. Jordão2, P. Evangelista2, E. Medina2, M. Forjaz Secca3; 1S. Neurorradiologia, H. Egas Moniz, Lisboa, PORTUGAL, 2Rm-sdi, Ressonância Magnética de Caselas, Lisboa, PORTUGAL, 3Physics Department, Fac. de Ciencias e Tecnologia / UNL, Lisboa, PORTUGAL. Introduction: MRI plays a decisive role in the discrimination of the epileptic focus. The aim is to identify a focal pathology, that can be surgically removed if patients become refractory to medication. Subtle abnormalities (mild-moderate hippocampal sclerosis - HS) may be difficult to interpret by qualitative MRI. In order to evaluate the effectiveness of quantitative methods, we compared the sensibility of the neuroradiological reading with a dual-quantitative analysis. Methods: 54 chronic temporal lobe epilepsy (TLE) patients undergoing pre-surgical evaluation with lateralization clearly identified were included. The qualitative protocol included a whole-brain assessment plus coronal 3D T1, FSE T2-weighted and FLAIR images acquired perpendicular to the main TL axis. HS criteria included abnormal T2 and FLAIR signal, atrophy on T1 and loss of the internal structure as definition. Quantitative MR data including T1-volumetry (HCVOL) and T2-relaxometry (HCT2) was processed using 34 age-matched control subjects [1]. Assessments of both studies were tested by comparing the qualitative diagnosis with the

S122

Epilepsy

standard deviations (SD) from the control mean of HCVOL and HCT2. Results: The ipsilateral HCVOL and HCT2 differed significantly between controls and patients. In 55% of the patients the HS diagnosis was concordant in both studies. HS was doubtful on the reading (HS?) in 11%, however both HCVOL and HCT2 were pathological. No signs of HS (no-HS) were observed visually in 14% of the patients with pathological HCT2. In 20% both studies were normal.

Table: Comparative results between HCVOL, HCT2 and neuroradiological reading. The table indicates how many SDs below the mean the HCVOL and HCT2 are in the majority of the patients in each group. In-between brackets are the values obtained in the minority of cases. Discussion: Abnormal signal intensity and morphology of the hippocampus may be difficult to detect visually in certain cases. Quantitative methods provide an objective means of assessing pathological changes and should be used in combination with qualitative studies to enhance the sensitivity. In our study visual assessment detected HS in the majority of patients. Hippocampal volumetry and T2-mapping provided localization in an additional small number of patients. References: 1. Gonçalves Pereira PM, et al [2002]Epilepsia 43(Suppl 8): 163. Support: Fundação para a Ciência e a Tecnologia BD 18498/98 and Fundação Grünenthal

222 Sexual dismorphism in volumetry of hippocampus M. A. Bertoni, N. E. Sclavi; Radiology, La Sagrada Familia, Bahia Blanca, ARGENTINA. Purpose: To assess normal volumetric values for the hippocampus related to sex and age and also differences between sexes. Methods: 109 patients with no history of seizures and normal electroencephalogram were examined with a standardized MR protocol including 3D acquisitions from which reformatted images at the longest head-tail axis of hippocampi were obtained. Contours of each hippocampus were traced manually and volume calculations were tabulated and analyzed with a forward stepwise multiple regression technique considering sex and age as independent and hippocampal volume as dependent variable for both males and females. Dilatations of ventricular system, congenital anomalies, disorders of neuronal migration, mass effect and atrophy were ruled out. All 109 clinical records were checked after a period of 5 years to rule out neurological and/or psychological disorders diagnosed subsequently to the moment at which the MRIs were obtained. Results: In the multiple regression analyses the following values were obtained: R=0.9335, R2=0.8716; adjusted R2=0.8704; F(2,215)=729.749; p>0.000001. Variances were similar for both sexes [FI(102,112)=0.798971<1.63] while beta values were different in males than in females [tII(214)=3.8459>1.980] obtaining, then a different equation for each sex.

Conclusion: Sex and age in our series appear to be related to the volume of hippocampus. Different regression parameters were found in male and female subjects when related to age.

223 Neuroradiological applications of hyperecho-TSE's at 3T: first clinical results M. Weigel1, S. Ziyeh2, J. Weber2, O. Speck1, J. Hennig1; 1Section of Medical Physics, Department of Diagnostic Radiology, University Hospital of Freiburg, Freiburg, GERMANY, 2Section of Neuroradiology, Neurocenter, University Hospital of Freiburg, Freiburg, GERMANY. Purpose: Multi-spinecho-sequences such as TSE are probably the most supporting pillar of diagnostic MRI due to their insensitivity to field inhomogeneity effects and direct T1- and/or T2-weighted contrast behavior. One major problem of TSE-sequences is their high RF power deposition that can be subdued by RF pulse schemes like hyperechoes[1] and TRAPS[2] (hyperTSE-sequences). HyperTSE-sequences proved to achieve similar contrast and SNR in normal controls compared to conventional TSE180°. However, their sensitivity to display pathologies still has to be evaluated in clinical studies. This preliminary study demonstrates that hyperTSE-sequences show both, an equivalent capability of resolving pathologies and possibly an even better contrast to noise ratio between cerebral WM and GM. Subjects and Methods: All experiments were performed on a 3.0T whole-body imaging system (Siemens Trio, Erlangen, Germany). A common TSE-sequence (TE/TR=100/5200ms, ETL=19-21, MTX=512x416, in-plane resolution≈(0.4mm)2, slth=2mm) was employed, in which the “TRAPS”-mechanism was implemented. Optimized sinusoidal flip angle variations were used such that the flip angles were ramped down from 180° to 60° in the outer portions of k-space. All acquisition times were approximately 4:30min. Results: Fig.1 demonstrates an example of high resolution T2weighted hyperTSE-images of a male patient (38y) with a presumed temporal ganglioglioma. The hyperTSE-images clearly identify and characterize the tumor and also display a very good contrast to noise ratio between WM, GM, and CSF. Fig.2 presents a comparison between a conventional TSE180° and the novel

Kidney and Prostate: Imaging

S123

hyperTSE-sequence for display of a focal cortical dysplasia. Both images show an equivalent capability of resolving the lesion. The conventional images show a brighter CSF, but the hyperTSE images display a better contrast between WM and GM. In Fig.3 a magnified hyperTSE-image of a male patient (23y) with periventricular leukomalica is demonstrated. The hyperTSE-image shows good contrast between WM, GM, and the hyperintense WM lesions. The SAR of the hyperTSE images presented here was reduced to approximately 33% of the equivalent TSE180°sequences. Discussion and Conclusion: As a preliminary result it was demonstrated that hyperTSE-sequences and conventional TSE180° have an equivalent capability of resolving different types of pathologies. HyperTSE-images even seem to show an intrinsinc higher contrast to noise ratio between WM and GM. In addition, RF power deposition can be drastically reduced by as much as 67%. This reserve in SAR allows more slices per acquisition and a higher ETL for shorter acquisition times. 1. Hennig J, Scheffler K, MRM 46:6-12 2. Hennig J, Weigel M, Scheffler K, MRM 49:527-35

Clinical Focus Session Kidney and Prostate: Imaging 3:00 pm - 4:00 pm

Jurriaanse

224 MRI accuracy in detection of the Wilms' tumor preoperative chemotherapy efficiency V. J. Dombrovsky; Radiology, Rostov State Medical Univ., Rostov-on-Don, RUSSIAN FEDERATION. Purpose: The magnetic resonance imaging (MRI) accuracy in the control of Wilms' tumor (WT) preoperative chemotherapy (PCT) efficiency was investigated. Subjects and Methods: The MRI and histopathological (including immunohistochemical findings) data of 53 patients, 24 boys and 29 girls, with a mean age of 3,4 ± 0,2 years (range 0,6 – 9 years), with proven WT were analyzed. All children underwent PCT (SIOP Protocol N 9). MRI was performed on 0,5 T unit (Tomikon S 50 Avance, “Bruker”, Ettlingen, Germany). Spin-, gradient-echo sequences, including fast, were used. Comparisons [coefficients of correlation (r) and association (r1); p < 0,05) were created using variance analysis. Results: The WT dimensions were measured before and after PCT in three sectional planes according to the greatest tumor’s linear sizes. The mathematical function “image algebra” was used additionally for correct estimation. High positive correlation level between MRI final data and pathologic findings, concerning WT dimensions (r = 0,94) and tumor structure secondary alterations (hemorrhage – r1 = 0,89; cysts – r1 = 0,92; necrosis – r1 = 0,85), was found. According to this result, we suggest to use a simple formula [1] for calculating the “reduction index” (RI), truly reflecting PCT effectiveness: RI = (X1 + X2 + X3)/ (Y1 + Y2 + Y3) [1],

S124

Kidney and Prostate: Imaging

X1-3 - is the greatest tumor’s size on the three sectional planes after PCT; Y1-3 - is the initial greatest tumor’s size on the three sectional planes. In 31 cases the RI value was 0,29 – 0,57. It means a considerable decrease of WT dimensions with marked tumor structure secondary alterations. In others (n = 22) – > 0,6 with moderate tumor tissue alterations. The second PCT course was required in last cases. Conclusion: MRI data are confirmed to be accurate for reflecting the true pathologic status of WT and reduction of this lesion after PCH.

225 Diffusion-weighted MR imaging in the evaluation of renal lesions E. Squillaci1, M. Cova2, G. Manenti1, M. Di Roma1, R. Pozzi Mucelli2, G. Simonetti1; 1Radiology, University of Rome Tor Vergata, Rome, ITALY, 2Radiology, University of Trieste, Trieste, ITALY. Purpose: To evaluate the capability and the reliability of diffusionweighted MR imaging (DWI) in the evaluation of normal kidney and different renal lesions. Subjects and Methods: 10 normal volunteers and 40 patients with known renal lesions underwent MRI of the kidneys by using a 1.5 T superconductive magnet. Axial fat suppressed TSE T2- and coronal FFE T1- or TSE T1-weighted images were acquired in each patient. DWI were obtained in the axial plane during breath-hold (17s) with a SE EPI single shot sequence (TR=2883, TE=61, flip angle=90°), with b value of 500 s/mm2. Sixteen slices were produced with slice thickness of 7 mm and interslice gap of 1 mm. ADC map was obtained at each slice position. The ADC was measured in an approximately 1 cm ROI within the normal renal parenchyma, the detected renal lesions and the collecting system if dilated. Results: ADC values in normal renal parenchyma ranged from 1.72 ± 0.15 x 10-3 mm2/s to 2.65 ± 0.34 x 10-3 mm2/s, while ADC values in simple cysts (n=13) were higher (2.87 ± 0.06 x 10-3 mm2/s to 4.00 ± 0.00 x 10-3 mm2/s). In hydronephrotic kidneys (n=7) the ADC values of renal pelvis ranged from 3.39 ± 0.48 x 103 mm2/s to 4.00 ± 0.00 x 10-3 mm2/s. In cases of pyonephrosis (n=5) ADC values of the renal pelvis lower than those of renal pelvis of hydronephrotic kidneys were shown (0.77 ± 0.17 x 10-3 mm2/s to 1.21 ± 0.13 x 10-3 mm2/s). ADC values of solid benign and malignant renal tumors (n=15) ranged from 1.28 ± 0.11 x 10-3 mm2/s to 2.23 ± 0.43 x 10-3 mm2/s. There was no overlap of ADC values of the normal renal parenchyma, renal pelvis of pyonephrotic kidneys, and solid renal tumors, except for the ADC values of one renal cell carcinoma that overlapped with the ADC values of normal renal parenchyma. Largely overlapping ADC values were found for simple renal cysts and renal pelvis of hydronephrotic kidneys, which showed significantly higher ADC values compared to the other patients group. Conclusion: DWI of the kidney seems to be reliable to differentiate normal renal parenchyma and different renal diseases.

226 Magnetic Resonance Urography (MRU). Is a full bladder helpful? J. A. C. Colville1, R. P. M. Killeen2, T. J. Geoghegan1, G. Hurley1, F. Regan1, S. Hamilton1, W. C. Torreggiani1; 1Radiology, Adelaide and Meath Hospital Incorporating the National Children's hospital, Dublin, IRELAND, 2Medicine, Adelaide and Meath Hospital Incorporating the National Children's hospital, Dublin, IRELAND. Purpose/Introduction: The purpose of our study was to evaluate if a full bladder improves visualisation of the renal tract during magnetic resonance urography (MRU). Subjects and Methods: 20 volunteers (10 male & 10 female) aged between 20-45 years were included in the study. The patients were scanned in a 1.5 Telsa Siemens MRI scanner using unenhanced HASTE sequences. Half of the patients had MRU initially with empty bladders and then returned with "full" bladders. The second group presented with full bladders, were scanned and after which they voided and were rescanned with empty bladders. All patients were adequately hydrated prior to MRU. Two independent consultant radiologists reviewed the PACS images, grading both right and left kidney and ureters according to their visibility. Results: Bladder filling prior to scanning improved the visualisation of the urinary tract with "full" bladder systems having 73.8% visualisation (+/- 3.34%). Empty bladders had 54% visualisation (+/- 2.1%). Concordance value between the radiologists was 81%. Same day studies showed a greater difference in the percentage of visualisation of the empty systems than those performed on separate days. Discussion/Conclusion: When feasible MRU studies should be performed with a full bladder to optimise visualisation of the renal tract.

227 High-resolution MR imaging for nodal staging in rectal cancer: Are there any criteria in addition to the size? J. Kim1,2, R. F. A. Vliegen1, J. M. A. van Engelshoven1, M. Kim2, G. L. Beets3, M. F. von Meyenfeldt3, R. G. H. BeetsTan1; 1Radiology, University Hospital of Maastricht, Maastricht, NETHERLANDS, 2Radiology, Yonsei University College of Medicine, Seoul, REPUBLIC OF KOREA, 3Surgery, University Hospital of Maastricht, Maastricht, NETHERLANDS. Purpose: MR Staging of nodal metastases in patients with rectal cancer using criteria based on size, shape and signal intensity can be difficult, because ≥50% of the nodes are less than 5 mm in size. Therefore new MR criteria were evaluated to see whether it can improve the MR assessment of nodal metastases in rectal cancer patients. Materials and Methods: From Jan 1998 and Dec 2000, 99 patients with primary rectal carcinoma underwent 1.5-T high-resolution MRI with a quadrature phased array coil. Sequences used were 2D T2W FSE and Gadolinium enhanced 2D T1W FSE with an in plane resolution of 0.78 × 0.78 and 0.39 × 0.39, respectively. Among them, 75 patients who had undergone TME were enrolled in this study. An MR radiologist, blinded for the histological results, randomly recorded the characteristics of each detectable node (LN); common criteria such as short diameter, the ratio of long to

Kidney and Prostate: Imaging short diameter and signal intensity on each sequence. New criteria such as the margin (smooth, nodular, spiculated, indistinct), a homogenous or mottled heterogeneous appearance, gross enhancement and its pattern, the distance of LN to the mesorectal fascia, the venous encasement, and the perirectal dirty fat signal. Results: Among 75 patients, 22 (29%) were node-positive. All patients who did not have detectable LN on MR were node-free (n=15). Presence of LNs ≥ 5mm was significantly higher in the node-positive group. Presence of LNs ≥ 9mm was seen only in the node-positive group. A spiculated border (a sensitivity of 45%, a specificity of 100%), an indistinct border (a sensitivity of 36%, a specificity of 100%), and a mottled heterogeneity (a sensitivity of 50%, a specificity of 95%) were correlated with LN positivity (p<.001, respectively). A venous encasement (n=4) and a dirty perirectal fat signal (n=3) were also significant (p<.05, respectively). The ratio of long to short diameter, signal intensity on T2w or T1w images, geographic heterogeneity, gross enhancement, and the distance of LN to the mesorectal fascia were not correlated with the nodal status. Conclusion: In addition to size, new criteria such as a spiculated or indistinct border, a mottled heterogeneous appearance, venous encasement and dirty perirectal fat sign could be useful to predict regional lymph node involvement in patients with rectal cancer.

228 T2 versus T3 for staging of prostate cancer: clinical value of the combination of pelvic phased-array coil (PPA) and integrated endorectal pelvic phased array coil (ERC-PPA) W. Pegios1, W. Bentas2, T. Vogl3, G. Bongartz1, W. Steinbrich1; 1Radiology, Kantonspital Basel, Basel, SWITZERLAND, 2Urology, Kantonspital Münsterlingen, Münsterlingen, SWITZERLAND, 3Diagnostic and interventional Radiology, Universitätsklinikum Frankfurt, Frankfurt, GERMANY. Purpose: Evaluation of the diagnostic value of the combined endorectal-body-phased-array technique regarding the staging of prostate carcinomas, especially in the defferentiation between T2 and T3. Methods and Materials: 54 patients with biopsy proven prostate carcinoma were examined on a 1.5 T scanner (Siemens, Symphony) with multiplanar orientations prior to radical prostatectomy. T2-w TSE (axial, cor) and T2-w FSE (axial) sequences were obtained unenhanced. After application of 0.2 mmol/kg bw GdDTPA, T1-w GRE was carried out using dynamic MRI. All images were prospectively interpreted by two observers without any knowledge of histology. Results: Observer 1 reached a sensivity of 100% and a specifity of 87% regarding the detection of extracapsulary growth and seminal vesicle infiltration (T2 versus T3). The overall staging accuracy amounted to 94%, the positive predictive value (PPV) to 90% and the negative predictive value (NPV) to 100%. In two cases with a histologically proven stadium pT2b, obersever 1 had diagnosed stadium pT3a. The MRI did not lead to an understaging, regarding the differentiation of pT2 and pT3, in any case by observer 1. The results of Obersever 2 were marginally better since he only overstaged one case, histologically proven pT2b, as pT3a. Therefore the sensivity and specivity accounted to 100% and 93%, respectively. Furthermore: overall staging accuracy 97%, PPV 95% and NPV 100%. Overall, the staging accuracy regarding all tumor stages was correct in 44 of 54 cases, which leads to a percentage of 81. The dynamic MRI showed no improvement regarding specificity (62%)in the differentiation between stages T2 and T3.

S125

Conclusion: The combined endorectal-body-phased-array technique play a significant role, in the preoperative staging of the prostate carcinoma. However, the differentiation between capsulary infiltration and penetration as well as the judgement of the seminal blatter occur to be difficult.

229 Verification of conformal radiotherapy doses using MRI G. P. Liney1, A. Heathcote2, A. W. Beavis3; 1Centre for MRI, Hull Royal Infirmary, Hull, UNITED KINGDOM, 2Physical Sciences, University of Hull, Hull, UNITED KINGDOM, 3Radiation Physics, Princess Royal Hospital, Hull, UNITED KINGDOM. Introduction: Advanced methods of delivering radiotherapy (e.g. IMRT) permit extremely conformal doses to be targeted to the tumour whilst sparing surrounding tissue. However these approaches, using dynamically changing beams, render conventional two-dimensional methods of verifying dose (e.g. x-ray film) inadequate. Tissue equivalent gels may be used which can be imaged with MRI to demonstrate radiation-induced changes in relaxation time. Previous gels have been shown to be unstable, difficult to manufacture or of limited accuracy, which has precluded their clinical use. A novel gel medium is described here based on Fong et al [1], which offers advantages over previous types of dosimeter. A R2-dose calibration has been established and this has been used to verify delivered intent in a known 3D dose distribution with good accuracy and at high resolution. Methods: Batches of gel were manufactured and irradiated incrementally between 0 and 30 Gy using a Varian 600C series linear accelerator. MR studies were performed on a 1.5 Tesla GE Signa scanner using a conventional head coil. Two dual-echo FSE sequences (TE/TR = 30,105 & 60,180 /4000 ms, ETL = 12) were acquired through the gel containers (25 cm FOV, 256 × 256 matrix) using a 5 mm slice thickness. A calibration plot of R2 versus dose was obtained and this was used to verify the dose distribution delivered to a large flask of gel from the same batch. This flask had previously been irradiated with 3 cm wide pairs of opposing beams to produce a cross of 5 and 10 Gy doses. Results: Figure 1 shows calibration plots of R2 versus dose for different compositions of gel together (M3, M9, M12) with previous results using BANG gel [2] as a comparison (P). The new gel demonstrates an extended range of linearity and a concentration dependent sensitivity (slope:intercept value). Figure 2 illustrates a dose map acquired in the large flask demonstrating the 0, 5, 10 and 15 Gy regions (scale in Gy). Values for each particular region agreed to within a few percent of delivered intent. Conclusions: This work demonstrates the efficacy of using MRI to provide absolute dose quantification at high spatial resolution in three dimensions. The technique will be important in the quality assurance and acceptance testing of IMRT. References: 1. PM Fong et al, DOSGEL 2001; 95, ISBN: 1 86435 549 2. 2. GP Liney et al, MAGMA 2002; 15(S1): 480.

S126

Angiography: Imaging and Technique structions and also to quantify relative changes in veins diameters. Each vein was segmented separately to get integral signal intensity over the investigated volume. The projection pixel value represents the size of the vein multiplied by MR signal intensity. After normalization using average MR signal from reference ROI, profiles perpendicular to the main venous direction were measured in 3 places equidistantly positioned from atrium border. Integrating the profile we got a value representing relative extent of cross-sectional area. Because of various contrast agent filling and dynamics, we used only relative values of CSA compared to the most proximal section. Comparison of relative CSA between MRA examinations are presented as results. Group 33 patients were examined in ablation preparation, 9 patients before and immediately after the ablation, 4 of them also after 3 months. Results: All but one examination were technically successful and provided relevant and useful anatomical information. After ablation, 4 clear stenos having CSA below 50 % were found and all these stenoses were stable also after 3 months. But we also found diffuse decrease of CSA down to ~60-80% on next 13 veins. Schematic situation for 9 patients is shown in fig.1.

Clinical Focus Session Angiography: Imaging and Technique 3:00 pm - 4:00 pm

Ruys

230 Pulmonary veins anatomy monitoring before and after RF ablation using a dynamic contrast enhanced MR angiography J. Tintera1, R. Cihak2, P. Fendrych1, E. Rolencova1, H. Mlcochova2, J. Kautzner2; 1Department of Radiology, IKEM, Prague, CZECH REPUBLIC, 2Department of Cardiology, IKEM, Prague, CZECH REPUBLIC. Introduction: Recently, the RF ablation intervention became to be an efficient method to treat atrium fibrillations. Contrast enhanced (CE) MRA was already shown as useful tool offering valid information about pulmonary veins anatomy for ablation planning. We use dynamic CE MRA to compare size of pulmonary veins before and after the ablation and to monitor possible changes in vessel cross-sectional area (CSA) in long time follow-up. Subjects and Methods: CE MRA measurements were done on Siemens Vision 1.5T with 3D GE FLASH sequence (TE=1.2ms, TR=3.2ms, α=90°). Volume of 30 slices (THK=3.4mm) was scanned 3-4 times with simultaneous application of contrast agent (20ml of Multihance, 3ml/s). Interpolated spatial resolution was 1.4x1.7x1.7 mm, time resolution 6-11 s per measurement. Time interpolation and subtraction of dynamically measured data was used to suppress pulmonary arteries and static tissue together with best veins visualization. Data segmentation allowed 3D recon-

Discussion: CE MRA is an ideal tool to visualize and control pulmonary vein anatomy before and after the RF ablation. Our data suggest that in more cases of insignificant decrease of CSA due to the ablation there is a tendency of reparation (in 5 from 8 veins). However, all significant stenoses remained stable after 3 months. Study was supported: IGA MZCR NA/7381-3

231 MRA of spinal arteries in TAAA-patients R. J. Nijenhuis1,2, W. H. Backes1, T. Leiner1, J. T. Wilmink1, M. J. H. M. Jacobs2, J. M. A. van Engelshoven1; 1Radiology, Maastricht University Hospital, Maastricht, NETHERLANDS, 2Vascular Surgery, Maastricht University Hospital, Maastricht, NETHERLANDS. Introduction: Surgical repair of a thoraco-abdominal aortic aneurysm (TAAA) may sometimes lead to paraplegia, due to spinal cord ischemia. Therefore pre-surgical visualisation of the spinal vasculature is of value. Recently Yamada1 has shown that with contrast enhanced (CE)-MRA it is possible to visualise spinal vasculature, instead of using X-ray angiography. Using MRA it is an enormous challenge to detect very small spinal arteries in a large body region. The most

Angiography: Imaging and Technique important vascular structure that should be visualised is the Adamkiewicz’ artery (AKA). In surgical literature it is stressed that the segmental artery (SA) which connects to the AKA should whenever possible be reattached when a prosthetic aorta is implanted. The purpose of this study was to investigate to which extent it was possible to consistently visualise the AKA in patients with TAAA using CE-MRA. Subjects and Methods: Phantom-study A phantom containing 10 thin tubes (0.13-6 mm) was used to find the smallest visible diameter. The tubes were filled with the contrast agent (15 mM GdDTPA) and were surrounded by a tissue mimicking fluid (T1 800 ms). Patient-study A phased-array spine coil was used on a 1.5 T system. FOV covered thoracic vertebra 5 (T5) till lumbar vertebra 5 (L5) Scanning parameters were: TR/TE/FA/pixel 5.9ms/1.9ms/30°/0.8x0.8 mm2. Sagittal slices were 0.6 mm thick. After bolus tracking, the subsequent CE-MRA sequence consisted of two dynamic phases each less than 40 s. The scans of 13 patients were analyzed with multiplanar reformations (MPR) and maximum intensity projections (MIP). Only vessels observed in the early phase and fading in the late phase were considered to be arterial. Results: Phantom-study The smallest diameter that could be distinguished from the surrounding tissue was 0.5 mm. Patient-study In all patients the AKA, the anterior spinal artery (ASA) and the connecting SA could be visualized (figure 1). Conclusion and Discussion: Using CE-MRA it is possible to detect the arteries feeding the spinal cord in TAAA patients. Anatomical studies2 have demonstrated that the diameter of the AKA varies from 0.5-1.5 mm, which falls in the range of our demonstrated 1.5T imaging capabilities. We suggest that this presurgical information is of sufficient quality to plan and guide surgery in TAAA patients. References: 1. Yamada N [2000] JCAT 24:362-368. 2. Koshino T [1999] J. Thorac. Cardiovac. Surg. 117:898-905.

S127

232 Contrast-enhanced MR angiography of the renal arteries: blinded multicenter crossover comparison of 0.1 mmol/kg gadobenate dimeglumine (Gd-BOPTA) and 0.2 mmol/kg gadopentetate dimeglumine (Gd-DTPA) M. Prokop1, M. Goyen2, P. Douek3, A. Vanzulli4, G. Schneider5; 1Dept. of Radiology, University Medical Center Utrecht, Utrecht, NETHERLANDS, 2Dept. of Radiology, University Hospital Essen, Essen, GERMANY, 3Dept. of Radiology, Hôpital Louis Pradel, Lyon, FRANCE, 4Dept. of Radiology, Ospedale Maggiore Niguarda, Milano, ITALY, 5Dept. of Radiology, University Hospital Homburg/Saar, Homburg/Saar, GERMANY. Purpose: To intraindividually compare 0.1 mmol/kg Gd-BOPTA (T1 relaxivity in blood=9.7 mmol-1s-1) and 0.2 mmol/kg GdDTPA (4.9 mmol-1s-1) for contrast-enhanced MRA of the renal arteries. Methods and Materials: 34 patients at 3 centers underwent two identical renal MRA exams at 1.5 T separated by >48 hours but <12 days. The randomized order for CM administration was GdBOPTA/Gd-DTPA in 18 patients and Gd-DTPA/Gd-BOPTA in 16 patients. A phase-encoded 3D-spoiled breath-hold sequence (TR/TE/FA=<5/<2/45°, FOV=32-36 cm, 2 mm slices (n=32), 1x1x2 mm spatial resolution and acquisition time=<30 sec) was used after test bolus injection and calculation of time delay. Two blinded, independent readers qualitatively assessed randomized subtracted MIP images from each exam for diagnostic quality. A 3point scale (diagnostic information poor=0, moderate=1, adequate=2) was applied to each of 9 vessel segments covering the abdominal aorta and left and right renal arteries to give an overall quality score between 0 and 18. Quantitative assessment (vessel SNR, vessel-muscle CNR) of unsubtracted source images was performed at ROIs placed on the supra-, juxta-, and infrarenal aorta and psoas muscle. Results: No significant qualitative differences between 0.1 mmol/kg Gd-BOPTA and 0.2 mmol/kg Gd-DTPA were noted by either reader (reader 1: Gd-BOPTA=15.15, Gd-DTPA=15.23; p=0.94; reader 2: Gd-BOPTA=16.77, Gd-DTPA=17.01; p=0.46). The order of treatments likewise produced no differences: readers 1 and 2 reported quality scores of 14.4±4.2 and 16.7±2.3, respectively, when Gd-BOPTA was the first CM, and 15.2±1.8 and 16.6±1.6, respectively, when Gd-DTPA was the first CM. The scores when Gd-BOPTA and Gd-DTPA were the second agents were 16.0±3.0 and 15.3±2.9, respectively (reader 1), and 16.9±2.0 and 17.4±1.4, respectively (reader 2). Quantitative evaluation revealed no differences in mean SNR and CNR although increasing SNR and CNR on descending the aorta was seen with Gd-BOPTA (Gd-DTPA: SNR=38/37/41, CNR=34/34/36; Gd-BOPTA: SNR=32/42/48, CNR=34/38/44, supra/juxta/infrarenal aorta, respectively). Conclusion: 0.1 mmol/kg Gd-BOPTA is at least equivalent to 0.2 mmol/kg Gd-DTPA for renal MRA.

233

Figure 1. MPR of the spinal arteries: AKA (arrowhead), ASA (large arrow) and the connecting SA (small arrow).

Optimisation of Contrast-Enhanced 3 step MRA of the lower extremities by infragenual compression H. G. Heidecker1, D. Bilecen1, M. Aschwanden2, K. Jaeger2, G. Bongartz1; 1University Institute for Diagnostic Radiology, Kantonsspital Basel Stadt, Basel, SWITZERLAND, 2University Institute for Angiology, Kantonsspital Basel Stadt, Basel, SWITZERLAND.

S128

Angiography: Imaging and Technique

Introduction: Quality of Contrast-Enhanced 3 step MRAngiography of the lower extremities is sometimes hampered by venous overlay of the crural arteries (Fig1)[1,2]. In PAOD-patients infragenual cuff compression of the lower extremities during normal MRA workup significantly reduces crural venous overlay and thus augments MRA imaging. Material and Methods: 53 Patients with Peripheral Vascular Disease (PVD) were retrospectively investigated. The compression group, randomly chosen, had a standard blood-pressure-cuff infragenually applied to one leg (40-60 mmHg) 3 minutes prior measurements. Patients were placed supine, foot first in MR-Scanner. A peripheral-surface-coil was used for detection of the signal. Patients received a standard care-bolus-triggered ContrastEnhanced-three-step-MRA of the lower extremities. Standard GdDTPA contrast agent was injected in the left antecubital vein with 1.5 ml/s followed by 15 ml saline flush. All coronal MRA data were acquired with a 3D-spoiled-gradientecho-sequence (matrix size: 512 x 134, TR: 4.45ms, TE: 1.4 ms, flip angle: 25°, TA: 27 s, FOV: 450 x 197 mm2, partitions 112, slice thickness: 1.6 mm). Venous contaminations of the crural arteries were evaluated for each leg separately by three radiologists. Venous contamination scores were 0 = no venous contamination, 1 = minor venous contamination, 2 = major venous contamination, 3 = no diagnostic value. Results: Table 1 shows evaluation of venous contamination. The control group shows venous overlay of crural arteries from no to minor (0-1) in 59% and major or more (2-3) in 40% for the left side. Similar results were observed for the right side. With infragenual compression venous contamination of crural arteries is no to minor in 93% and major or more in only 7%. These data are resembled by the Mean Venous Contamination Score (MVCS). MVCS is roughly 1.52 for both legs in the control group and 1.64 for the uncompressed side of the compression group. MVCS is 0.79 for the compressed legs in the compression group. Discussion: Findings show that infragenual compression reduces venous contamination of the crural arteries significantly. We could show, that infragenual compression can be implied in routine CEMRA workup of the lower extremities. We therefore suggest this easily applicable technique to be used for all patients undergoing CE-MRA of the legs. 1. Wang Y et al. [2002] Invest Radiol 37(8):458-63 2. Bilecen et al., Subdiastolic compression technique for optimized infrapopliteal contrast-enhanced MRA and its effect on contrast agent distribution, Submitted 2003

234 MSCT and fast 3d MR angiography in preoperative evaluation of vascularized fibular harvest: Work in progress A. Magenta Biasina1, L. Caverni1, C. Bonifacio1,2, G. Pompili1, M. Goisis3, G. Cornalba1,2, F. Biglioli3; 1Institute of Radiology, San Paolo Hospital, Milan, ITALY, 2University of Milan, San Paolo Hospital, Milan, ITALY, 3Maxillo-facial Surgery, San Paolo Hospital, Milan, ITALY. Purpose: To evaluate the role of routine preoperative MSCT and MR angiography of lower extremity arteries in patients planned to have a free fibula harvested. Methods and Materials: We examined 30 patients with lower extremities vascular imaging, 20 underwent MSCT angiography and

Angiography: Imaging and Technique 10 fast 3D MR angiography. In all of these patients normal dorsalis pedis and posterior tibial pulses were palpable. Only the patients without vascular abnormalities underwent surgery. Vascular images were compared to surgical vascular anatomy. Results: One patient was ruled out from surgery because the MR angiography showed bilateral anterior tibial arteries absence. Two patients were ruled out based upon the RM angiography which showed bilateral posterior tibial artery absence. In the remaining 27 patients MSCT and MR angiography demonstrated normal vascular anatomy of lower extremity arteries and the fibula was harvested. There was 100% accuracy between imaging studies and surgery findings in the depiction of vascular anatomy. In our experience MSCT angiography has a better role in showing the vascular calcifications and better spatial resolution than MR angiography, but the technique is more time-consuming in the postprocessing and The patients are exposed to ionizing radiation and to the risk of allergic reaction to iodate contrast agent. Fast 3D gadolinium-enhanced MR angiography has lower biological costs and doesn’t need postprocessing time. Conclusion: The blood supply of the fibular flap is based in the peroneal artery, for this reason it’s essential to determine preoperatively whether the adequate perfusion to donor extremity will persist after sacrifice of the peroneal pedicle.It is essential to identify the inability of free fibula trasplantation due to an ostruction or an hypoplasia of the peroneal artery. In our experience we actually use Fast 3D MR angiography in the presurgical planning of vascularized fibula harvest.

235 Subdiastolic compression technique for optimised contrastenhanced MRA of the hand D. Bilecen1, H. G. Heidecker1, M. Aschwanden2, K. Jaeger2, G. Bongartz1; 1University Institute for Diagnostic Radiology, Kantonsspital Basel Stadt, Basel, SWITZERLAND, 2University Institute for Angiology, Kantonsspital Basel Stadt, Basel, SWITZERLAND. Introduction: Diagnostic value of bolus chase contrast-enhanced magnetic resonance angiography (CE-MRA) of the hands is frequently hampered by venous contamination [1,2]. We suggest an easy to apply cuff compression technique at the upper arm to significantly reduce venous contamination of hand angiographies. Material and Methods: Six healthy volunteers were included. Three minutes prior measuring a blood pressure cuff was unilaterally applied to one upper arm and inflated to 50 mm Hg. Contrast agent (Dotarem 0.2 ml/kg body weight) was administered through the contralateral antecubital vein followed by saline flush. MRA acquisition: coronal, 3D spoiled GRE sequence, TR 4.5 ms, TE 1.3 ms, TA 20 s, matrix size: 448 x 176, FOV of 320 x 230 mm2, partitions 56, slice thickness 1 mm. Six consecutive measurements were acquired, the first serving as mask. Based on MIPimages venous contamination was scored and ranked by two radiologists. Wilcoxon’s signed rank test was used for statistical analysis. Results: Figure 1 shows no venous contamination of the right hand, where upper arm compression was applied. The interpretation of the angiogramm of the left hand, where no cuff compression was applied, is impaired, due to significant venous contamination. Figure 2 displays the scoring of mean venous contamination over five consecutive MIP-images. Venous contamination increases from MIP 1-5 in both hands, and is overall lower on the compressed than on the uncompressed side. Difference in venous over-

S129

lay of MIP 4–5 is not significant (p> 0.05). The difference in venous contamination from MIP 1-3 is significant (p < 0.05). Discussion: Subdiastolic cuff compression of the upper arm can be used to augment CE-MRA of the hand by significantly reducing venous contamination. The mechanism is still under discussion but is presumably due to a backpressure mechanism [3,4]. Since venous overlay is delayed by upper arm compression, ultrafast scanning is not necessary. The impact of this technique on pathological conditions is subject of further investigation. 1. Goldfarb J.W. et al. [2001] AJR 177:177-1182 2. Wentz KU et al. [2003] Lancet 361:49-50 3. Bernstein EF [1993] Vascular Diagnosis 4th ed: Mosby 4. Bilecen D et al., submitted March 2003

S130

Perfusion

Poster Walking Tour Perfusion 3:00 pm - 4:00 pm

236 A simulation comparing three methods for perfusion quantification by bolus tracking F. Risse, J. M. Boese, S. Aumann, L. R. Schad; Abt. Biophysik und Medizinische Strahlenphysik, Deutsches Krebsforschungszentrum, Heidelberg, GERMANY. Introduction: Perfusion measurements by bolus tracking have achieved a great importance in studies examining stroke patients, but are also used in other organs e.g. the kidneys. In addition to singular value decomposition (SVD) [1], there are some other methods in use for perfusion quantification. In this work, deconvolution using orthogonal polynomials (DOP) [2] and the ‘gradient-relationship method’ (GRM) [3] are compared to SVD. Methods: Concentration time courses were simulated by convolving a gamma-variate as the arterial input function (AIF) with an exponential and a box as the residue function (R) [1]. Regional blood flow (RBF) values of 30, 60, 90 ml/min/100ml were chosen to simulate brain tissue and values of 200, 400, 600 ml/min/100ml to simulate renal cortex. Relative regional blood volume (rRBV) was kept constant at 3% and 5% for brain and at 25% and 30% for kidneys. Gaussian distributed noise was added after conversion of concentration to signal time courses according to different signalto-noise ratios (SNR = 10, …, 70). RBF was calculated after backconversion of the curves using SVD, DOP and GRM. 1024 simulations were performed for each combination of parameters. Results: RBF is underestimated by all methods independently of rRBV. Relative differences between fitted and true flow up to 25% (SVD) and 50% (DOP, GRM) were found for mismatching conditions. In general, SVD showed the lowest differences, whereas DOP (exponential R) showed slightly higher errors for short mean transit times. Certainly the differences were strongly dependent on the choice of the threshold (SVD) resp. the number of polynomials (DOP). The underestimations of RBF by GRM were generally more than 20% higher than for SVD and DOP (Fig. 1). Discussion: All methods resulted in an underestimation of RBF for noisy data. However, SVD makes the best estimation of RBF for all combinations of true RBF, RBV and SNR. Improved results can be achieved by optimising deconvolution methods with respect to the SNR. Although SVD showed the best results, the DOP approach might also be useful at bolus passages with longer transit times. Despite the strong underestimation of RBF by GRM, it can be useful for qualitative perfusion analysis because of its simplicity and independence of recirculation, since only the initial slope of the concentration curve is used. References: 1. Østergaard L et al, Magn Reson Med 1996;36:715-725 2. Stritzke P et al, IEEE Trans Med Imaging 1990;9:11-23 3. Miles KA, Br J Radiol 1991;64:409-412

237 FAIR true-FISP perfusion-weighted imaging of the kidneys at 3 T: Comparison with 1.5 T P. Martirosian1, U. Klose2, H. Graf1, F. Schick1; 1Department of Diagnostic Radiology, Section on Experimental Radiology, University of Tübingen, Tübingen, GERMANY, 2Department of Neuroradiology, Section on Experimental NMR of the CNS, University of Tübingen, Tübingen, GERMANY. Introduction: In addition to a generally higher signal-to-noise ratio (SNR) a longer T1 relaxation time contributes to a relative increase of the perfusion related signal yield by performing the arterial spin labelling (ASL) studies at higher fields. The combination of ASL techniques with imaging sequences insensitive to susceptibility effects might provide useful applications concerning internal organs and the musculoskeletal system. The FAIR True-FISP technique eliminates both, susceptibility artifacts and image distortions commonly observed in EPI images. The goals of this study were to evaluate the performance of the FAIR True-FISP method at 3.0 T and to compare the results to those obtained at 1.5 T on the human kidneys. Subjects and Methods: The examinations of the kidneys of healthy volunteers were performed on a 1.5 T and a 3.0 T whole body MRI systems (Siemens Medical Solutions, Erlangen, Germany). The FAIR True-FISP sequences were implemented for perfusion studies with TR=4.6 ms, TE=2.3 ms, TI=1200 ms, matrix=128×128, and FoV=360×360 mm2. A 10.24 ms FOCI adiabatic inversion RF pulse with µ=5, β=935 was used to obtain steep slopes for slice-selective inversion. The recorded slice thickness was 8 mm, the inversion slab thickness was 20 mm. Single slice studies were performed both in breath-hold and under normal respiration. The relative perfusion signal change δS=(Ssel-Sns)/Sns in the whole kidneys and SNR=(Ssel-Sns)/σ noise in the cortex were calculated to compare the perfusion-weighted images at 3.0 and 1.5 T (Ssel, Sns are the signal intensities in selective and non-selective inversion images and σnoise is the noise standard deviation in the subtraction image). Results: Figure 1 shows a comparison of FAIR True-FISP images acquired from a healthy volunteer. The images clearly show that the cortex is well perfused, whereas the medulla is nearly lacking in any perfusion signal. The 3.0 T image (left) shows significantly better image quality. A quantitative comparison is shown in Table 1. SNR in the cortex is improved by a factor of ~2.0 at 3.0 T. The 3.0 T results also show an increase of the relative signal change of ~100 %.

Perfusion

Tab. 1. The relative perfusion signal change and SNR in the kidneys at 1.5 and 3.0 T. Field Right Kidney Left Kidney Strenght δS (%) δS (%) (T) SNR SNR 1.5 8.8 ± 3.3 10.6 8.9 ± 3.4 8.9 3.0 20.6 ± 7.5 20.5 18.1 ± 7.3 16.9

Discussion: The presented work conforms the FAIR True-FISP imaging technique to be suitable for perfusion imaging on the kidneys at 3.0 T. The short TR and TE of the sequence leads to a marked reduction of susceptibility artifacts. Our preliminary results clearly indicate significant improvement in SNR and sensitivity of the perfusion signal going from 1.5 to 3.0 T.

238 Glomerular filtration rate: measurement with dynamic contrast-enhanced MRI and cortical-compartmental model in the rabbit's kidney L. Annet1, L. Hermoye1, F. Peeters1, F. Jamar2, J. Dehoux3, B. E. Van Beers1; 1Radiology, St-Luc University Hospital, Brussels, BELGIUM, 2Nuclear Medicine, St-Luc University Hospital, Brussels, BELGIUM, 3Laboratory of Experimental Surgery, Université Catholique de Louvain, Brussels, BELGIUM. Introduction: The purpose of this study is to describe and validate the use of MRI and a cortical-compartmental model to measure glomerular filtration rate (GFR) in the rabbit's kidney. This model takes tracer output from the renal cortex during the sampling time into account, as well as delay and dispersion of tracer input between the aorta and the renal glomeruli. We compared the corticalcompartmental model with the known Patlak model where these elements are not included in the fitting. Materials and Methods: Dynamic, fast gradient-echo MR images of the kidney were obtained in 10 rabbits after bolus injection of 0.05 mmol kg-1 of gadoterate dimeglumine. Regions of interest were manually drawn in the aorta, the renal cortex, and around the entire right kidney in order to obtain intensity versus time curves. After conversion into concentration versus time curves, the GFR was calculated by fitting the curves in the renal cortex with the cortical-compartmental model, according to the following equation: CROI(t)=fa C'a(t)+k21 ∫C'a(y) e-k12(t-y)dy, where CROI represents the concentration of the contrast agent in the kidney, C'a, the concentration in the plasma of the renal cortex, fa, the fractional blood volume, k21, the GFR, and k12, the output of the renal tubules. The GFR was also calculated with the Patlak model in the whole kidney and in the cortex, according to the following equation: CROI(t)/Ca(t)=fa+k21 ∫Ca(t)dt/Ca(t), where Ca is the concentration in the aortic plasma. The plasma clearance of chromium-51-EDTA (Cr-51-EDTA) was used as a reference method to calculate the GFR. Results: A significant correlation was obtained between the GFR calculated with the cortical-compartment model and the clearance of Cr-51-EDTA (r = 0.821, p = 0.004). In contrast, no significant correlation was found between the GFR calculated with the Patlak

S131

model in the whole kidney and the Cr-51-EDTA clearance (r = 0.628, p = 0.052). The correlation between the GFR calculated with the Patlak model in the cortex and the Cr-51-EDTA clearance was significant (r = 0.744, p = 0.014), but the GFR calculated with this cortical-Patlak model (0.41 ± 0.20 mL min-1 kg) were significantly lower than the GFR calculated with the cortical-compartmental model (0.68 ± 0.31 mL min-1 kg) (p = 0.005). Conclusion: The cortical-compartment model is valid to measure glomerular filtration at MRI and should be preferred to the Patlak model.

239 High-resolution myocardial perfusion mapping in small animals in vivo F. Kober, I. Iltis, M. Izquierdo, M. Desrois, D. Ibarrola, P. J. Cozzone, M. Bernard; Centre de Résonance Magnétique Biologique et Médicale UMR CNRS 6612, Faculté de Médecine, Marseille, FRANCE. Robust in vivo methods adapted to the measurement of myocardial perfusion in rats and mice still remain a challenge. In this work, we propose a gradient-echo spin-labeling technique as an alternative to previously used inversion-recovery snapshot-FLASH (1, 2). The main advantages of this approach are high sensitivity at high spatial resolutions and the ability to measure myocardial perfusion in freely breathing animals in vivo at very high heart rates. An ECG-gated inversion-recovery gradient-echo technique with synchronization to respiration was used to acquire T1 maps. In this approach, only one phase-encoding step per inversion pulse is acquired at each cardiac cycle. The inversion is also ECG-gated. Inversion pulse and acquisition of a set of 32 gradient-echoes are triggered after a completed respiration cycle. T1 maps were obtained with slice-selective and global inversion. The two T1 maps are related to each other via perfusion. The relation used by Belle et al. (1) was used to calculate perfusion maps. The sequence was applied to seven freely breathing Wistar-Kyoto rats under isoflourane anaesthesia at 4.7T. The measurement was repeated under dobutamine stress for four animals. The in-plane spatial resolution was 234 x 468 um². At a mean heart rate of 350 bpm, this lead to typical measurement times of 25 minutes. The figure shows a typical perfusion map obtained from a healthy rat. Perfusion values above 20 ml/g/min and below -20 ml/g/min were masked in order to exclude large vessels. The high spatial resolution avoids partial-volume effects at the interface between myocardium and intraventricular blood. The mean perfusion values (group average ± SD) were 5.5 ± 0.7 ml/g/min.(rest, n=7) and 11.1 ± 1.9 ml/g/min (stress, n=4). The rest values are slightly higher than those previously observed (1, 2). We attribute this difference to the use of isoflurane instead of pentobarbital. Gradient-echo inversion-recovery proves to be robust and particularly useful for the study of small rodents at high fields. This work was supported by the "Imagerie du petit animal" program (2002-2003). 1. Belle V, Kahler E, Waller C, Rommel E, Voll S, Hiller K, Bauer W, Haase A. [1998] J Magn Reson Imaging; 8: 1240-5. 2. Waller C, Kahler E, Hiller KH, Hu K, Nahrendorf M, Voll S, Haase A, Ertl G, Bauer WR. [2000] Radiology; 215: 189-97.

S132

Perfusion 1. Neyran, MRM 2002;48:166-179 2. Larsson, MRM 2001;46:272-281 3. Epstein, MRM 2002;47:482-491 Regional myocardial blood flow in a normal pig SI R1 rMBF(ml/100g/min) rMBF(ml/100g/min) Rest 224+-240 157+-175 Stress 1224+-727 600+-385

Microsphere rMBF(ml/100g/min) 89 583

240 In vivo parametric myocardial perfusion mapping S. Carme, E. Canet-Soulas, B. Neyran; Creatis, INSA Lyon, Villeurbanne, FRANCE. Introduction: First pass contrast enhanced Magnetic Resonance Imaging (MRI) has been proven able to assess myocardial perfusion. Using an intravascular contrast agent CMD-A2-Gd-DOTA (P717, Guerbet), we have presented in a previous study the robustness of deconvolution methods to assess physiological parameter of perfusion in an isolated pig heart preparation (1). The present study propose to evaluate the validity of our method in the in-vivo case, when transferred to an experimental setting close to the clinical conditions. Parametric pixel maps of Regional Myocardial Blood flow (rMBF) computed on MR signal intensity (SI) are presented. The conversion of SI into the tissue relaxation rate R1 limit the saturation effect and improve the perfusion parameters computed on R1 curve. Results are compared to radioactive microspheres reference blood flow. Methods: First-pass images after P717 bolus injection were obtained with a 1.5 Tesla whole-body MR scanner, using a saturationrecovery Turbo-Flash sequence. Experiments are conducted under rest and stress in normal and ischemic pigs. The signal response limitation implies to convert the SI into R1 using the one compartment model defined for the use of the intra-vascular contrast agent in fast water exchange assumption (2). The percent enhancement (PE) (3) is a good index to measure the enhancement of both perfusion kinetics of SI and R1. rMBF is computed on SI and R1 kinetics using the ARMA deconvolution technique defined (1). Results are presented in pixel maps and by mean +- standard deviation measured in the lateral and septal myocardial sectors. Results: The PE in the blood is 174% in SI and 266% in R1. In the normal myocardium, the PE is 45% in SI and 55% in R1. Maps of rMBF, computed with SI and R1, are presented at stress (Figure). The mean and standard deviation are given at rest and under stress in normal myocardium (Table). In ischemic area, mean rMBF is always negatives and can be interpreted as close to zero. Discussion: Conversion of SI into R1 enhances the perfusion kinetics and then improves the deconvolution process. RMBF values identified on R1 curves are closer to the reference microsphere myocardial blood flow (table). Figure shows the ability of rMBF maps to assess the local myocardial viability in in-vivo case. However, the estimation process should take into account the neighborhood of each myocardial pixel to reduce the standard deviation of estimated rMBF.

241 Effect of combretastatin A-4 on rat bladder tumors implanted on nude mice: preliminary longitudinal study by anatomical and carbogen induced functional MRI C. D. Thomas1, C. Walczak1, J. Kaffy2, J. Jouanneau3, A. Volk1; 1U350 inserm, Curie Institute, Orsay, FRANCE, 2Umr176 cnrs, Curie Institute, Paris, FRANCE, 3Umr144 cnrs, Curie Institute, Paris, FRANCE. Purpose/Introduction: Tubulin disrupting agents like combretastatin A-4 exhibit potent anti-angiogenic and anti-vascular activity in malignant tumors. The purpose of the current preliminary study was to assess how tumor growth and fMRI signal upon carbogen breathing were affected by administration of combretastatin A-4 disodium phosphate (CA-4P). Subjects and Methods: NBTII-FGF1 rat bladder tumors were implanted subcutaneously in the flank of female nude mice. CA-4P was i.p. injected (100mg/kg in 0.02ml/g Nacl 0.9%). Experiments were performed on a Bruker Biospec (4.7T). MRI acquisitions were synchronized with respiratory motion. The imaging protocol comprised a fast spin echo sequence for high resolution screening of the whole tumor and measurement of growth curves (FOV 3x3x2.5 cm3, matrix 256x256x50, TR about 2s, TEw=41ms) and a fast gradient echo sequence, allowing an entire T2* weighted image acquisition per respiratory cycle, to perform "BOLD"-type imaging (FOV=(3cm)2, matrix 128x80, Slth 1mm, TR/TE=26.6/16.3ms, a=30º). 80 sets of 15 contiguous slices comprising the tumor were acquired during the following protocol of breathing gas administration: image set 1-7: air, image set 8-48: carbogen, image set 49-80: air. Tumor fractions with signal enhancement under carbogen breathing (T+) were determined by counting corresponding voxels in manually defined ROIs covering the whole tumor. 6 tumors were studied and followed every day during 2 to 4 days with one daily CA-4P injection. FMRI images were acquired before and about 30min after CA-4P injection. Results: CA-4P considerably slowed down tumor growth for all the tumors. Before treatment T+ varied between tumors. We observed a decrease in T+ quickly after CA-4P injection (about 35 minutes). In addition, after histology examination (Hematoxylin stain), we noted the presence of pre-necrotic and necrotic regions corresponding to regions which exhibited positive response to carbogen before CA-4P and no or even negative response afterwards.

Perfusion Discussion/Conclusion: These preliminary results indicate that growth of experimental bladder tumor is slowed down by CA-4P treatment. Histology results confirmed in our model the tumor structure modification and destruction by this agent. Furthermore, the fMRI analysis showed, rapidly after CA-4P injection, a decrease in T+, compatible with a decrease in tumor perfusion. Therefore, in this study, a decrease of the tumor fraction characterized by a signal enhancement under carbogen breathing seems to be an early indicator of drug action. Acknowledgments: This work was supported by the Curie Institute (tumor angiogenesis program) and by the CNRS-INSERM interdisciplinary program "Imagerie du Petit Animal"

242 Value of dynamic contrast enhancement MRI for assessing the disease activity of multiple myeloma: a comparative study with histology and clinical markers S. Nosàs1, T. Moehler2, F. Kiessling1, K. Wasser1, M. Heilmann3, R. Bartl4, I. Zuna5, S. Delorme1; 1Division of Oncological Diagnostics and Therapy, DKFZ (German Cancer Research Center), Heidelberg, GERMANY, 2Department of Hematology/Oncology/ Reumatology, University of Heidelberg, Heidelberg, GERMANY, 3Division of Biophysics and Medical Radiation Physics, DKFZ (German Cancer Research Center), Heidelberg, GERMANY, 4Medizinische Klinik III, Klinikum Großhadern , Ludwig- Maximilians- Universität München., München, GERMANY, 5Clinical Cooperation Unit Radiotherapeutical Oncology, DKFZ (German Cancer Research Center), Heidelberg, GERMANY. Purpose: Dynamic contrast-enhanced MRI (dMRI) has been used for assessing treatment response in multiple Myeloma (MM) of the spine [1]. However, it is so far unproven that increased contrast uptake in bone marrow indeed reflects tumor infiltration. Therefore, aim of this study was to examine histologically whether increased contrast enhancement in bone marrow of patients with MM corresponds with local tumor infiltration and increased vessel density. Methods: The spina iliaca superior posterior of 24 patients with MM was examined using a dMRI protocol with a pump-controlled infusion of Gd-DTPA. After dMRI, a biopsy of the spina iliaca superior posterior was taken. Using a two-compartment model [2] the MR parameters amplitude (reflecting plasma volume) and kep (exchange rate constant) on the region of the biopsy were calculated and compared with the histological and clinical data. Results: kep was higher in patients with tumor involvement than in those without tumor involvement at the biopsy site (p<0.05). Both dMRI parameters (A and kep), were higher in lesions with a marked infiltration than in mild infiltration at histology (p=0.01). The amplitude, but not kep, was higher in lesions with a high vessel density score than in those with a low vessel density at histology (p=0.01). Conclusion: Areas of increased contrast uptake in the bone marrow of MM patients very probably reflect tumor involvement. Furthermore, the degree of contrast enhancement is correlated with the degree of both tumor infiltration and vessel density. References: 1. Moehler TM, Hawighorst H, Neben K, Egerer G, Hillengass J, Max R, Benner A, Ho AD, van Kaick G, Goldschmidt H [2001] Int.J.Cancer 93:862-8. 2. Tofts PS [1997] J.Magn.Reson.Imaging 7:91-101.

S133

243 Dynamic estimation of R1 and R2* in normal and tumour bearing mouse livers by gradient-echo multi-echo MRI during bolus administration of a paramagnetic contrast agent S. Chileg1, J. L. Dimicoli1, C. D. Thomas1, C. Walczak1, M. J. Plessis2, F. Perin2, A. Volk1; 1Inserm u.350, Curie Institute, Orsay, FRANCE, 2Genotoxicity and Carcinogenesis, Curie Institute, Orsay, FRANCE. Purpose: The purpose of the work was to implement and evaluate in vivo, in normal and tumour bearing mouse livers, a protocol based on gradient-echo multi-echo (GEME) MRI for simultaneous dynamic estimation of R1 and R2* relaxation rates at 4.7 Tesla during contrast agent (CA) administration. Indeed, in addition to separation of R1 and R2* effects of the CA, the relation between ∆R1 and ∆R2* may yield information related to microvascular architecture [1]. Subjects and Methods: Experiments were performed at 4.7 T on a BRUKER Biospec. Normal (n=3) and tumour bearing livers (n=2, chemical induction by 5,9-dimethyl-7H-dibenzo[c,g]carbazole) were studied in XVIInc/z female mice. The respiratory triggered imaging protocols comprised MSME imaging (tumour localisation/characterization), a series of GEME images with varying excitation angle α (pre-contrast R1 measurement), dynamic GEME imaging during bolus administration of Dotarem® (one image/respiratory cycle (≈2sec), matrix 128x64, FOV 3x2.4cm2, α = 30°, TR/TE = 25.5ms/(2.7 ; 5.4 ; 8.1 ; 10,8; 13.5; 16.2)ms). Dotarem® (0.5M) was injected via the tail vein using a power injector (40µl @ 240µl/min). Pre-contrast and dynamic R1 and R2* were computed in liver and tumour ROIs using home-written IDL® code. Dynamic R1(t) estimations were obtained from TE=0 extrapolated signal and pre-contrast R1 and R2* as described before for excised perfused liver [2]. Relationship between ∆R1(t) and ∆R2*(t) was evaluated by linear regression analysis. Results: Upon CA administration, rapid increase of R1 and R2* was observed in the 3 control liver ROIs, 2 apparently normal liver ROIs of tumour bearing livers, and 1 tumour ROI (hepatocellular carcinoma (HCC)). Three other tumour ROIs (attributed to hypovascular cholangioma (n=2) and necrosis (n=1)) didn't exhibit substantial relaxation rate increase. ∆R1(t) and ∆R2*(t) values were highly correlated in control liver ROIs (R2 = 0.94, mean slope: d(∆R1)/d(∆R2*) = 0.020±0.004), in tumour bearing liver ROIs (R2(ROI1)=0.86, slope(ROI1)=0.030±0.001; R2(ROI2)=0.89, slope(ROI2)=0.039±0.001)), and the HCC-ROI (R2=0.84, slope=0.020±0.001). Discussion/Conclusion: This study focussed on the relationship between ∆R1(t) and ∆R2*(t) obtained with an original protocol based on respiratory triggered rapid contrast enhanced GEME imaging. Preliminary results suggest a higher slope in tumour bearing liver ROIs with respect to control livers, which might be related to changes in tissue/microvascular architecture since this affects R2*. Acknowledgements: This work was supported by the Curie Institute (tumour angiogenesis program) and by the CNRS-INSERM interdisciplinary program "Small Animal Imaging" References: 1. Kiselev VG [2001] Magn. Reson. Med 46:1113-1122 2. Dimicoli JL, Patry J, Volk A [2002] MAGMA 15(S1):284

S134

Perfusion

244 Susceptibility weighted imaging of brain masses D. Haddar1, V. Sehgal2, M. E. Haacke3, Z. DelProposto4, A. Sloan5, L. Zamorano5, W. Kupsky6, J. R. Reichenbach7, K. Tong8; 1Radiologie, Hôpital Jean Verdier, Bondy, FRANCE, 2MRI Radiology Department, Harper University Hospital, Detroit, MI, 3MRI Institute for Biomedical Research, MR Research Facility, Detroit, MI, 4Radiology, Harper University Hospital, Detroit, MI, 5Neurological Surgery Department, Harper University Hospital, Detroit, MI, 6Department of Pathology, Harper University Hospital, Detroit, MI, 7Institute of Diagnostic and Interventional Radiology, Friedrich-Schiller University, Jena, GERMANY, 8Departments of Radiology and Radiation Medicine, Loma Linda University, Loma Linda, CA. Purpose: To evaluate the contribution of susceptibility weighted imaging (SWI) in the study of brain masses. Materials and Methods: SWI is a high resolution, three-dimensional, fully velocity compensated gradient echo sequence. Postprocessing is applied using both magnitude and phase images to increase the conspicuity of the veins and other sources of susceptibility effects. This sequence has been applied to 44 patients (24 males, 20 females, aged 15 to 89, mean 50.3 years) with brain masses, pre-and/or post-contrast and compared to conventional sequences (T1, T1 post-contrast, T2, proton density, FLAIR and diffusion weighted imaging). Seventeen cases had pathological correlation. Seventeen cases were primary brain tumors, 19 were metastasis and 8 other diagnosis. Results: T1 post-contrast was only slightly better than SWI pre- or post-contrast for tumor detection and delineation. SWI was more sensitive for showing blood products and venous vasculature. SWI showed a useful FLAIR-like contrast and added useful complementary information to conventional T1 post-contrast sequences for internal architecture study of the lesions. Good pathologic correlations were found for blood products, slightly lower for venous vasculature. Conclusion: SWI may prove useful for tumor characterization by its ability to show blood products and venous vasculature.

245 Discretising the convolution product in DSC-MRI based quantification of CBF S. Sourbron1, I. Stubbe2, R. Luypaert1, M. Osteaux1; 1Befy, Vrije Universiteit Brussel, Brussels, BELGIUM, 2Agel/math, Université Catholique de Louvain, Louvain La Neuve, BELGIUM. Purpose/Introduction: The quantification of Cerebral Blood Flow (CBF) with Dynamic Susceptibility Contrast (DSC) enhanced MRI starts from the model C=CBF×CA*R, where C is the tracer concentration, CA is the Arterial Input Function (AIF), R is the residue function, and * denotes convolution [1]. The equation is solved for the unknown X=CBF×R after which CBF is found as the maximum of X. Since the data are measured at times t0,t1,…,tn-1 with interval TR, the convolution is discretised by linear interpolation and written as a matrix equation C=TR×AX. Using the notations introduced in Eq.[1], row i of A is given in Ref.[1] by Eq.[2]. Our calculations however lead to Eq.[3] for i≠ 0 and Ai=0 for i=0, which introduces a correction to Eq.[2] in the first column and on the diagonal of A. In this work we demonstrate the relevance of this correction by simulating the perfusion model for Grey Matter. Methods: The AIF was modeled as a gamma-variate [1] with

exponent α =3.0, characteristic time β =1.5s and bolus arrival time 5s. The residue function was modeled as an exponential with Mean Transit Time of 4s and CBF=60ml/100ml/min. We simulated the convolution for various values of TR and performed a deconvolution at TR=1.5s, regularizing the problem by Truncated Singular Value Decomposition with a cut-off selected using Generalized Cross Validation [2]. No noise was added to the data to expose the error induced by discretisation. Results: Fig.[1] depicts the convolution C for 0.05s
Processing and Quantification: Imaging

Poster Walking Tour Processing and Quantification: Imaging

S135

multi-resolution approach and deformation spheres with decreasing radius (the model’s deformability increases). Results: It takes on average one hour for a model with 85,000 vertices to deform (Pentium4, 2GHz). The automatic method has been successfully tested and validated on four different pelvis and femurs with the same parameters: deformability (radius from 2cm to 1cm), number of iterations (10000), number of landmarks (12) and number of resolutions (3). The error between manual and automatic segmentation is less than 3% of the total number of voxels for bones. Conclusion: We demonstrate a robust method to achieve a general, shape-independent reconstruction anatomical modelling of human bones from 3D MRI datasets that minimises manual tasks. As a potential application, orthopaedists can use these models for both preoperative planning and postoperative guides. Joint functional modelling, which requires precise soft tissue anatomical models is under investigation within the framework of our project.

3:00 pm - 4:00 pm

246 A methodology for anatomical 3D modelling of patient’s bones from MRI B. Gilles1, L. Yahia-Cherif1, R. Perrin2, L. Moccozet1, N. Magnenat-Thalmann1, J. Vallée2, F. Terrier2; 1Miralab, University of Geneva, Geneva, SWITZERLAND, 2Département de Radiologie, Hôpitaux Universitaires de Genève, Geneva, SWITZERLAND. Introduction: This paper presents a methodology for anatomical modelling of human bones from MRI. Bone-specific 3D acquisition protocols are defined. Generic models are reconstructed and then used as deformable models to reconstruct any individual models automatically. As a first study, we focussed on femur and pelvis bones. The long term of our current project, funded by the CO-ME (Computer-aided and Image Guided Interventions) research body, is to provide a new set of tools for anatomical and functional modelling of the full leg in order to improve the success rate of orthopaedic surgeries. Methods: Image acquisition The pelvis and femur of healthy volunteers are imaged using a Philips Intera 1.5T MRI system. A spin echo sequence is defined with TR=578ms, TE=18ms, Matrix/FOV=512x512/40cm and slice thickness=2mm to 10mm. This protocol attains excellent contrast between cortical bone and muscles, which is critical for segmentation. Generic models reconstruction Segmentation is performed using a discrete snake procedure to extract the contours of bones on every slice. After rigid series registration, iso-surfaces are calculated to provide the anatomical shape. The segmentation is validated by visual inspection. Individualized models reconstruction First, the generic model is elastically initialised with a landmarkbased approach and Bookstein’s Thin-Plate-Splines interpolation. Then, the model is deformed automatically by optimisation of an energy function which is composed of an external energy term, measuring the matching between the model and image edges, and an internal energy term that maintains a smooth and connected model. In order to avoid convergence into local minima, we use a

247 Phase imaging and post processing in BOLD venography A. Rauscher1, J. R. Reichenbach1, M. Barth2, S. Witoszynskyj2, E. Moser2, W. A. Kaiser1; 1Institute for Diagnostic and Interventional Radiology, Friedrich-Schiller-University, Jena, GERMANY, 2Institute of Medical Physics, University of Vienna, Vienna, AUSTRIA. Purpose/Introduction: Recently, high resolution MR-venography utilizing the BOLD-effect of the MR-signal has been developed [1], which allows to visualize slow flowing blood or blood clots. Because BOLD venography uses phase images together with the corresponding magnitude images, problems due to phase wraps may arise with conventional homodyne demodulation post-processing schemes [2]. Here we use a new image processing method [3] which is based on phase unwrapping in image space and subsequent high-pass filtering of the unwrapped phase images to obtain improved MR-venograms. Subjects and Methods: High resolution, 3D-MR data were acquired on a 1.5T system (Magnetom Vision, Siemens, Germany) using a velocity-compensated gradient-echo sequence (TE=40ms, TR=67ms, α=25°, FOV=240×240×48 mm3, matrix=512×512×32). A region growing phase unwrapping algorithm was applied to the phase images [4]. The unwrapped phase images were high-pass filtered and scaled so that phase values related to veins are assigned values between 0 and 1, whereas phase values corresponding to parenchyma are set to 1. This ensures that upon multiplication of this phase mask with the magnitude image veins appear dark in the resulting venogram while the parenchyma remains unchanged. Results: Figs. 1a and 1b show the original phase image and the corresponding unwrapped image, respectively. Fig. 1c displays a

S136

Processing and Quantification: Imaging

mIP over seven slices of the unwrapped phase data set. Note the rich information in this image which shows small veins and larger anatomic structures (e.g., basal ganglia). The increased magnetic susceptibility in the basal ganglia (due to the high iron content) leads to phase changes and gives rise to superb image contrast. Different image information is obtained when combining phase and magnitude information in the venogram (Fig. 1d). In contrast to Fig. 1c the ventricles appear as dark structures in this projection image.

Fig 1. (a) Original phase image. (b) Corresponding unwrapped image. (c) mIP (7 slices) of the phase data-set. (d) Venogram computed from both the magnitude and phase images. Discussion/Conclusion: Evaluation of the vasculature of the human brain in both its normal and pathological state is currently gaining importance. Phase unwrapping and separate analysis of phase and magnitude images as well as their combination in venograms may yield valuable information about sources of contrast. References: 1. Reichenbach JR, Haacke EM [2001] NRM Biomed 14:453-467 2. Noll DD et al. [1991] IEEE Trans Med Ima 10:154-163 3. Rauscher A et al. [2003] JMRI in press 4. Xu W, Cumming I [1999] IEEE Trans Geosci Rem Sens 37:124-134

248 Unsupervised assessment of abdominal fat distribution by MRI V. Positano1, A. M. Sironi1, A. Gastaldelli1, R. Petz1, M. F. Santarelli1, L. Landini2, M. Lombardi1; 1MRI Lab, CNR Institute of Clinical Physiology, Pisa, ITALY, 2Dep. of Information Engineering, University of Pisa, Pisa, ITALY. Purpose: High levels of visceral adipose tissues (VAT) have been linked to an increased risk of several obesity-related illnesses. Magnetic resonance imaging (MRI) may provide a method to measure adipose tissue volumes safely and accurately. However, previous methods for assessment of VAT by MRI were laborious and time consuming and often limited to the analysis of bi-dimensional data. We describe and evaluate an automatic method for 3D assessment of the quantity and the distribution of adipose tissue by MRI. The accuracy of the automatic method was compared with manual approach. Subjects and Methods: Ten subjects underwent whole-abdomen MRI. Subjects were imaged on a GE Signa LX System 1.5T scanner using a body coil. 32 transverse T1 weighted 256x256 images (TR=135, TE=4.2, flip angle=90°, FOV= 50cm, pixel size 1.875mmX1.875mm) were acquired from each subject with a slice thickness of 3mm with no overlap. Collected data were transferred to dedicated workstation and analyzed by the developed automatic software and by a manual procedure. The proposed methodology uses a fuzzy clustering approach to segment the adipose tissue. First, a k-mean fuzzy algorithm is used in order to classify the image grey level distribution in three classes, i.e. air, fat tissue and

muscular tissue. After, a region grooving algorithm is used on the map describing the air distribution in order to exclude from the analysis the patients arms. The third step uses an active contour approach applied on fat tissue map to automatically define two ROIs, as depicted in figure (left): the first one is defined by external (1) and internal (2) boundaries of subcutaneous adipose tissue (SAT), the second one is defined as a rough region (3) that includes visceral fat. The first ROI is used to evaluate SAT. Statistical analysis of image histogram extracted from the second ROI allows the evaluation of the VAT. The method provides SAT and VAT volumes in unsupervised manner as well as VAT/SAT ratio.

Results: The mean processing times were about 60 minutes for manual analysis, about 10 minutes for automatic unsupervised method. Results of simple regression analysis are showed in figure (right). The correlation between manual and automatic measurements was good for both SAT (p<0.001, R2=0.96) and VAT (p=0.031, R2=0.68). Conclusion: The proposed method has a significant advantage over manual tracing in terms of speed. Moreover, unsupervised analysis was significantly correlated with the manual one providing effective 3D assessment of abdominal fat distribution in reasonable time.

249 Vessel size imaging at low contrast agent concentrations I. Troprès1, L. Lamalle1, R. Farion2, M. Décorps2, C. Rémy2; 1Irm 3t, CHU de Grenoble, Grenoble cedex, FRANCE, 2U594, INSERM, Grenoble cedex, FRANCE. MR vessel size imaging quantifies a vessel size index (VSI) from measurements of the variations in transverse relaxation rates R2 and R2* due to the injection of an intravascular susceptibility contrast agent1. The method has proven to be a relevant tool for rat brain tumour characterisation2. Its application to human brain is however not straightforward3. In a previous work1, it was shown that the analytical developments underlying the expression for the VSI value were valid in vivo on rats provided that the contrast agent dose was high enough. For AMI-227 (Sinerem®) at 2.35 T, a dose of 200 µmol Fe/kg was used. Currently, clinical examination allows only approximately four times lower doses. In view of this constraint, the dose dependence of VSI estimation was assessed from experiments on glioma bearing rats at lower contrast agent concentrations. The MRI experiments, described in ref. 1, were carried out at 2.35 T, 3 to 4 weeks after C6 glioma cell injection in the striatum of 7 male Wistar rats. Simultaneous measurements of R2* and R2 were performed before and after injection of AMI-227. Measurements were done at 3 different doses: 30, 110 and 205 µmol Fe/kg. Three regions of interest were defined on the MR images: in the peritumoral area (high contrast enhancement), in the tumour (necrosis excluded) and in the contralateral side. For each dose means and standard deviations were computed over the 7 rats. To assess the bias associated with the use of low contrast agent concentrations

Processing and Quantification: Imaging VSI obtained at 30 and 110 µmol Fe/kg was divided by VSI at 205 µmol Fe/kg. VSI at lower contrast agent concentrations was found overestimated with respect to the value obtained at 205 µmol Fe/kg. At 30 µmol Fe/kg VSI can be up to 2.7 times the value measured at the highest dose. With lower concentrations are also associated higher standard deviations. Overestimation of VSI with decreasing dose was predicted in reference 1. The ratios measured here are in good agreement with the simulations (figure 5, reference 1). For clinical protocols, the results indicate that the highest possible dose and/or higher magnetic fields should be preferred in order to increase the signal-to-noise ratio of R2 and R2* variations measurements. 1. Troprès I et al [2001] Magn Reson Med 45:397-408 2. Troprès I et al [2002] ESMRMB, abstract 41 3. Strecker R et al [2002] ESMRMB, abstract 40 Ratios of VSIs mean ± SD VSI30 / VSI205 VSI110 / VSI205

contralateral 2.7 ± 1.5 1.2 ± 0.4

intratumoral 1.9 ± 0.7 1.3 ± 0.4

peritumoral 2.6 ± 1.3 1.5 ± 0.3

S137

Biospec spectrometer. GE images (TR/TE=100/10ms) were acquired during air, air-CO2 and carbogen (95% O2) enrichment. The change induced by carbogen (VF) signifies vascular density and tissue perfusion. The change generated by CO2 (VD) corresponds to blood flow and volume. 70% partial hepatectomy (PHx) was performed on adult male Sprague-Dawley rats. Liver cirrhosis was induced by IP administration of TAA twice weekly for 8 weeks. Rats were scanned daily before and on days 0-10 after PHx (n=5) and weekly during the 8 weeks of TAA administration (n=4). Results: In healthy rats, liver VF values are 20 times (0.04±0.02) greater than those of muscle tissue (0.002±0.02)(p<0.01), suggesting a higher vascular density and blood content in the liver. Opposed to the muscle tissue (0.002±0.01), VD values of the liver are negative (-0.03±0.02)(p<0.01). CO2 enrichment causes vasoconstriction to the hepatic artery, hence liver deoxyhemoglobin level rises. During the cirrhotic process, VD values gradually turn positive (0.003±0.02) and VF values decrease (0.0006±0.007) (fig 1,2). These changes could indicate structural and functional alterations of sinusoids and portal pressure affecting hepatic hemodynamics.

250 MRI as a novel tool for monitoring hemodynamic changes in the regenerative and cirrhotic liver H. Harel1, E. Gros2, I. Matot3, G. Spira4, E. Galun5, I. Vlodavsky6, R. Abramovitch1; 1The Goldyne Savad Inst. for Gene Therapy, and MRI/MRS lab, HBRC, Hadassah Hebrew University Hospital, Jersualem, ISRAEL, 2Department of Pediatric Surgery, Hadassah Hebrew University Hospital, Jersualem, ISRAEL, 3Department of Anesthesiology and Critical Care Medicine, Hadassah Hebrew University Hospital, Jersualem, ISRAEL, 4Department of anatomy and cell biology, The Bruce Rappaport Faculty of Medicine, Rappaport Family I, Haifa, ISRAEL, 5The Goldyne Savad Inst. for Gene Therapy, Hadassah Hebrew University Hospital, Jersualem, ISRAEL, 6Department of Oncology Labs, Hadassah Hebrew University Hospital, Jersualem, ISRAEL. Introduction: Many liver diseases, as cirrhosis and cancer, involve hemodynamic changes. Currently, there are limited tools for monitoring these changes in the liver and they are invasive or ex-vivo. Angiogensis is the formation of new blood vessels. In liver regeneration, hepatocyte proliferation precedes the angiogenic process. Consequently, liver hemodynamics are immensely affected till regeneration concludes. In cirrhosis, reduced liver sinusoidal functionality results in altered hemodynamics. Previously1, we have demonstrated that changes in oxygen saturation, blood volume and blood flow can be detected by MRI using BOLD contrast. The aim of this study is to implement this method for monitoring changes in liver hemodynamics non-invasively. Methods: MRI experiments were preformed on a 4.7 T Bruker

Fig 1. Left column – before TAA administration, middle – after 1 week, right – after 4 weeks. Top row – SE images (TR/TE=521.5/ 17.9ms), middle – VF maps, bottom – VD maps.

S138

Processing and Quantification: Imaging shown. Note that the gray matter images accurately depict the deep gray structures (caudate, thalamus). Discussion/Conclusion: Tissue segmentation into gray matter, white matter, and CSF using multiple SS FSE images is important for quantification of brain metabolites. It has previously been demonstrated that accurate CSF determinations can be made using single shot FSE images (4), and the method presented here extends the technique to determine tissue composition as well. The method is particularly suitable for fast high field and clinical MRSI applications due to its short scan time and relatively simple processing of the data. References: 1. Pham DL et al. Annu Rev Biomed Eng 2:315-37 (2000). 2. Windham, JP et al. J Comput Assist Tomogr 12:1-9 (1988). 3. Jacobs MA et al. Brain Res 1-2:83-94 (1999). 4. Horska, A et al. Magn Reson Med 48:555-558 (2002). Supported by NIH P41 RR1524

Fig 2. Average VF and VD values during TAA administration.*TTest compared to Pre p<0.05 After Phx, VD and VF values change similarly (0.01±0.001, 0.009±0.007 respectively) and as regeneration proceeds they are gradually restored according to the delayed angiogenesis. Conclusion: Concisely, this MRI method is a new and non-invasive tool for following liver morphological and functional changes during regeneration and cirrhosis, and in future could be essential for monitoring liver disease and therapies. 1. Abramovitch R [1999] Cancer Res. 59:5012-5016.

251 A rapid method for brain tissue segmentation: Application to quantitative proton MR spectroscopic imaging A. Horska1, M. A. Jacobs, Ph.D.1, V. D. Calhoun, Ph.D.2, A. Arslanoglu1, P. B. Barker, D.Phil.1; 1Radiology, Johns Hopkins University, Baltimore, MD, 2Institute of Living (IOL), Yale University, Hartford, CT. Purpose and Introduction: Interpretation of quantitative proton MRSI data can be improved by knowledge of the voxel brain tissue composition. However, techniques for brain tissue segmentation (1), requiring the acquisition of high resolution MRI data, can be time consuming. Our aim was to develop a fast method, based on analysis of fast spin echo (FSE) MRI recorded at different echo and inversion times. Subjects and Methods: Six healthy subjects (6M, 20-35 years old) were examined at 1.5 and 3 T. The segmentation MR data set consisted of single shot (SS) FSE scans (TE=7/100/500 ms) and an inversion recovery (IR) SS-FSE (TI/TE=500/31 ms) performed at the same slice thickness (15 mm/4 slices) and locations as the MRSI sequence. The FSE scans were repeated with 3 mm slice thickness to obtain reference signal intensity vectors from pure gray matter, white matter, and CSF pixels. Scan time for each FSE sequence was approximately 10 seconds. Segmentation into gray matter, white matter, and CSF was performed with the Eigenimage (EI) filter technique (2,3). The EI is a linear filter that maximizes the projection of a desired tissue while it minimizes the projection of undesired tissues onto a composite image (eigenimage). Results: Image segmentation using the EI tool required approximately 10 minutes processing time. Examples of acquired MR images of one 15-mm slice (with a proton density (TE=7ms), T2 (TE =100ms/500 ms), and T1 contrast (TI/TE=500/31ms) and segmented gray matter, white matter, and CSF images obtained by EI are

252 Rapid total body fat quantification by magnetic resonance imaging A. de Greiff, F. Vogt, M. Ladd; Diagnostic and Interventional Radiology, Medical Facilities, University of Essen, Essen, GERMANY. Introduction: Body fat content is known to be related to a number of diseases. MRI has been shown to be an accurate method for determination and quantification of internal and subcutaneous body fat. The purpose of our study was to evaluate a rapid data acquisition technique as well as an appropriate image segmentation software which allows for minimal user interaction and merest inter observer variability. Methods: 12 healthy volunteers (mean age 28 years, range 18-36) with varying body mass indices (19-40 kg/m2) were examined. Different imaging techniques were evaluated (e.g. Flash, True FISP, and TSE) in order to find a compromise between acquisition time, spatial resolution and fat segmentation capability. All images were acquired using a conventional 1.5 T scanner (Siemens Sonata, Erlangen, Germany) equipped with a rolling table platform (MRInnovation, Essen, Germany). By pulling the subjects through the iso-center of the magnet multiple consecutive volumes were acquired delivering whole body data sets. Fat segmentation was carried out using a custom made software package utilizing a region growing algorithm with edge detection and signal inhomogeneity correction. Results: 80 to 90 images, depending on subjects’ size, from the arms to the legs were collected in the transversal plane. MR acquisition lasted from 3 to 12 minutes, depending on the sequence.

Processing and Quantification: Imaging The slice thickness was 10 mm separated by 10 mm gaps in each of the measurements. The signal to noise gain of the longer lasting sequences (TSE) was demolished by increasing motion artifacts, especially in the abdominal region. The best results regarding patient compliance, inter operator variability, and reproducibility were achieved utilizing a T1 weighted 2D-Flash sequence (TR 18 ms, TE 2.2 ms, FA 70°, FOV 50 cm, 205 x 256 matrix). The chosen echo time yielded to opposite fat and water signals so that magnitude and phase information contributed separately to the segmentation algorithm. MRI revealed total fat of 32-54 %, subcutaneous fat of 27- 44 % and internal fat of 5-9 %. The MRI results also demonstrated good correlation to bioelectrical impedance analysis (Impedance) and dual-energy X-ray absorptiometry (DEXA). Conclusion: The protocol of our study resulted in a noninvasive rapid quantification of the entire body fat, allowing for accurate assessment of internal and subcutaneous fat relationship and also different treatment regimens. Especially short examination duration of the favored sequence (T2-FLASH) and tolerable post-processing time make it a feasible procedure for clinical trials with numerous subjects, e.g. prevention studies.

253 Partial volume segmentation using grey level slope information D. C. Williamson, N. A. Thacker, S. R. Williams; Isbe, University of Manchester, Manchester, UNITED KINGDOM. Introduction: Segmentation of an image can be achieved using the grey-level distribution1. However, this is often found to be inaccurate even when supplemented by local smoothness assumptions. There remains a need for more accurate determination of a pixel’s tissue content. Previous studies make the assumption that the occurrence of partial volume pixels is negligible. However this assumption has been shown to be inadequate1. In our work, local information related to image structure is included directly, by modifying the Bayesian model using the local grey-level slope and the grey-level information. In examples where the grey-level information from two images is required to achieve 10% volumetric estimation accuracy2, the inclusion of the grey-level slope doubles the information available for partial-volume estimation. Method: From Bayes theory the probability, P(j|g,s), of classification j given grey level, g, and slope, s, can be rewritten as P(j)P(g,s|j)/Σj(P(j)P(g,s|j)). The likelihood, P(g,s|j), is written as P(g|j)P(s|g,j), where the P(g|j) is the likelihood of g given j and P(s|g,j) the likelihood of s given j and g. The grey-level distributions associated with a tissue were modelled as Gaussian distributions and those associated with tissue mixtures were modelled using a triangle-Gaussian convolution. P(g|j) was constructed as a combination of Gaussians and convolutions. The grey-level slope distribution was modelled using a combination of Rician distributions3; one for the distribution associated with noise and one for each tissue boundary. A 2D histogram (grey level and slope) was created and fitted to the model. The probabilities required for a Bayesian classifier were obtained from the model. Results: In a slice through a human leg (1a) three tissues are visible; bone, muscle and fat. The fat lies on the outside of the leg, in the bone marrow and in a phantom. Figure 1b shows the segmentation of muscle using only the grey level information. The muscle is reasonably well identified, however, it is quite clear that some pixels at the fat/bone and fat/air boundaries have intensities close to those of pure muscle and the grey-level segmentation can-

S139

not separate them. In contrast, inclusion of the slope information allows identification of the partial volume region (1c). Similar improvements were found when segmenting brain images. 1. K. W. Fleischer et al. IEEE Trans. Med. Img. 17:74-86 1998, N. A. Thacker et al. Brit. J. Rad., 74, 234-242, 2001. 2. M. Pokric, et al, proc. MIUA2001, 77-80, 2001 3. H. Gudbjartsson et al. Magn. Reson. Med., 34, 910-914, 1995.

254 Correction of intensity non-uniformity with wavelet smoothing function in MRI data S. T. Chung1, Y. Kim1, W. S. Ryu1, I. K. Hong2; 1Research & Development, Medinus, Yonginsi, REPUBLIC OF KOREA, 2Research & Development, Korea Polytechnic University, Seoul, REPUBLIC OF KOREA. Introduction: A bias field, non-uniformity of MRI image, is one of the effects that should be removed by post image processing procedure to reproduce uniform image. N3 method, which calculates the bias field of image, was introduced for reproducing more uniform image [1]. Originally Gaussian was used and this smoothing of bias field has significant impact on the performance of whole procedure [1]. The method suggested in this study is using wavelet function for smoothing and compared with Gaussian for how much improvement in uniformity can be achieved [2]. Subjects and Methods: For bias field smoothing, wavelet function rearranges the coefficients of base functions and the resulting first half of image size contains low frequency components. Leaving first half and force the rest half to be 0. This returns only high frequency components of bias field back to original image. The differences to be made in wavelet smoothing method are more iteration of smoothing process and proper scaling factor for resulting bias field intensity. This scaling factor takes a role to increase wavelet results up to the desired level. Following equations are applied to compare the uniformity and SNR of original, Gaussian, and wavelet images. Uniformity = (Smax – Smin)/(Smax + Smin) × 100 (%) SNR=(1/3)xSUM(signal mean/noise stdi), i=1, 2, 3 Results: In Figure 1, each center profile indicates original image, Gaussian smoothing N3, and wavelet smoothing N3. Wavelet smoothing result represents most uniform profile. In table, uniformity percentage represents wavelet smoothing method built the most uniform image among three and SNR also rose a little while the noise standard deviation of background stayed still.

(a) Original image (b) N3 with Gaussian (c) N3 with Wavelet Figure 1. Phantom Images

Processing and Quantification: Imaging

S140 Uniformity

Signal Mean

(%)

Noise Std.Dev. A

B

SNR C

Original

49.8

2692

7.41

8.08

7.96

344.9

Gaussian

40.5

2568

7.31

8.08

8.30

332.0

Wavelet

16.0

2950

7.96

7.85

7.80

367.1

Table. Uniformity & SNR of phantom image

(a) Original image (b) N3 with Gaussian (c) N3 with wavelet Figure 2. L-spine image Discussion/Conclusion: The property of wavelet to divide image’s high frequency and low frequency components helps to remove bias field from image and to improve uniformity of image. We concluded this wavelet smoothing method needs to be tested for various clinical cases and RF coils. References: 1. John G. Sled, et al. IEEE Transaction on Medical Imaging, 17(1), pp.87-97, 1998 2. John J. Benedetto, et al. Wavelets, pp.80, 457-460(1994)

255 Cortical thickness of the frontopolar area in typically developing children and adolescents S. O'Donnell1,2, M. D. Noseworthy3,4, B. Levine5,6,7, S. Blaser3,8, M. E. Brandt9, M. Dennis1,2,5; 1Brain and Behaviour Program, The Hospital for Sick Children, Toronto, ON, CANADA, 2Institute of Medical Science, University of Toronto, Toronto, ON, CANADA, 3Diagnostic Imaging, The Hospital for Sick Children, Toronto, ON, CANADA, 4Medical Biophysics, University of Toronto, Toronto, ON, CANADA, 5Department of Psychology, University of Toronto, Toronto, ON, CANADA, 6Medicine (Neurology), University of Toronto, Toronto, ON, CANADA, 7Rotman, Research Institute, Toronto, ON, CANADA, 8Medical Imaging, University of Toronto, Toronto, ON, CANADA, 9The University of Texas, Medical School, Houston, TX. Introduction: The frontopolar area of the brain is expanded in humans, relative to other species. This has led researchers to propose that evolutionary changes in the frontal polar cortex (FPC) were associated with the development of uniquely human behaviours. Our purpose is to plot the developmental change in cortical thickness of the frontopolar region across late childhood and adolescence, and to test for structure-function relationships between FPC thickness and ratings of executive function Subjects and Methods: FreeSurfer was used to rebuild structural MRIs and measure thickness of the cortical ribbon in 35 typically developing children and adolescents (age 8-20 years). Reconstructed models were linked to their original images using SUMA (Surface Mapping with AFNI), and Brodmann area (BA) labels created in standard space were mapped to the FreeSurfer models in order to isolate and average thickness measures of the labeled regions. Cortical thickness was calculated for the frontopolar area (BA 10) and 2 comparison regions: the dorsolateral prefrontal area (BA 46 and lateral BA 9) and the striatal area (BA 17). (Fig. 1) Behavior Rating Inventory of Executive Function (BRIEF) scores were used to assess executive function. The BRIEF allows for age-

and sex-normalized quantification of executive function, based on parent reports of everyday behaviour. There are 8 domains (Inhibit, Shift, Emotional Control, Initiate, Working Memory, Plan/Organize, Organization of Materials, and Monitor) that in combination produce three summary scores: a Behavioral Regulation Index (BRI), a Metacognitive Index (MI), and a Global Executive Composite (GEC). Results: There is a significant inverse association between age and FPC thickness (r= -0.591, p<0.0005), and DLPFC thickness (r= 0.495, p<0.005), such that cortical thickness decreases with age. There was no association between age and thickness in the striatal region. (Fig. 2) No effects of gender or laterality were found. Structure-function results reveal associations (p<0.05) between left hemisphere FPC thickness and BRIEF scores for Inhibit, Shift, Emotional Control, Working Memory, Plan/Organize, Monitor, BRI, MI, and GEC. No other associations existed. Discussion/Conclusion: Decline in FPC and DLPFC thickness with age is consistent with volumetric studies that reveal frontal gray matter loss during adolescence. It is believed that this loss is due to developmental synaptic pruning and increased myelination. Behavioural associations between the left FPC thickness and executive function are consistent with assertions that the FPC, uniquely expanded in humans, mediates higher-order behaviours.

Functional MRI: Applications

Scientific Session Functional MRI: Applications 4:30 pm - 6:10 pm

S141

257 Willem Burger

256 Age dependence of the BOLD response in children and adults W. Richter1, M. C. Richter2; 1Chemistry, Princeton University, Princeton, NJ, 2Center for the Study of Brain, Mind, and Behavior, Princeton University, Princeton, NJ. Introduction: The change in BOLD signal consequent to neuronal activity is only incompletely understood, but it is apparently dependent on each subject’s individual physiology. If groups of subjects are to be compared by means of their fMRI signal, it is essential to exclude systematic differences in the BOLD response between these groups, as these differences would constitute a source of statistical bias. We now tested the (null) hypothesis that there is no correlation between the typical timecourse of the BOLD signal and the subject’s age. This hypothesis has been tested before (1) but could not be significantly rejected in that study. For the first time, we included children in our study. Subjects and Methods: We acquired technically acceptable data from 30 subjects (mean age: 26 y, range 7-61 y), some of whom were imaged in two or three distinct sessions, for a total of 45 experiments. We presented a brief visual stimulus (500 ms of a flickering checkerboard) to subjects 15 times with a long interstimulus interval (25 s) while imaging at high temporal resolution (Bo=3 Tesla, TR = 294 ms, TE = 30 ms, matrix size 64x64, 5 slices through the occipital lobe). We identified activated pixels in the visual cortex in an independent experiment. Averaged Time courses were fitted with a sum of Gaussian functions. Results: In a parametric analysis, we found a significant correlation between the trailing edge of the response and age [p<0.001] for an ROI consisting of the ten most significantly activated voxels; we also found significant correlations for other ROIs. We furthermore divided our subject pool into three age groups; we found the trailing edge to be significantly earlier in the youngest group (7 to 19 years) than in the other two groups (see Figure). Discussion/Conclusion: Our results indicate that age-related changes in the timecourse of the hemodynamic response are a potential source of bias in experiments comparing different age groups, and that a comprehensive characterization of this response (for example, in areas of the brain other than sensory and motor cortices) as a function of age is necessary. This will be particularly important for pediatric studies. Reference: 1. D’Esposito M et al., NeuroImage 10, 6 (1999)

Executed and imagined movement studied with high temporal resolution at 3 Tesla C. Windischberger1, R. Cunnington2, E. Moser1,3; 1Institute of Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Howard Florey Institute, University of Melbourne, Melbourne, AUSTRALIA, 3Department of Radiodiagnostics, University of Vienna, Vienna, AUSTRIA. Introduction: Even simple movement tasks require a number of processes, including motor planning, preparation and execution. Part of these circuits have been shown to be active also for imagined movements [1]. However, controversy still remains whether the primary motor cortex (M1) is also part of the shared neural resources recruited [2]. In this study we have used event-related fMRI with very high temporal resolution to examine specifically the role of M1 and the supplementary motor area (SMA) in a precued, imagined movement task. Subjects and Methods: We have studied 8 healthy, right-handed subjects on a 3 Tesla Medspec S300 scanner (Bruker Biospin, Germany) with gradient-recalled EPI (MA=64x64, 4 slices, TE=40ms) using a repetition time of 300ms. Subjects were asked to perform a set of 12 executed and imagined finger to thumb movements each, triggered via an acoustic countdown. They were monitored for movement speed and precision. Data analysis was done in SPM using a finite impulse response approach with separate regressors for every three time points per trial period. Such only minimal assumptions on the actual haemodynamic response are required and it is possible to generate activation maps in steps of 900ms. From the activation maps and the corresponding anatomical images, we defined ROIs for M1 and SMA and calculated mean time courses, both for executed and imagined movements. Results: The figure shows time courses of (a)SMA and (b)M1 activity in a single subject. Start of the (imagined) movement is indicated by a dotted line. Signal intensity during execution trials is plotted with solid lines, during movement imagination with dashed lines. SMA activation is increasing constantly after trial onset, reflecting the BOLD counterpart of the Bereitschafts- or readiness potential. Signal in the SMA is significantly higher for the executed movements throughout the trial period, while activity in M1 is increased after the movement onset only. Although these effects are generally reproduced in all subjects, we have found considerable inter-subject variability in the haemodynamic response. Discussion: Our results indicate that there is only minimal involvement of the primary motor cortex for imagined movements. Variability found between subjects has also been shown very recently [3] and might be caused by individually differing movement planning strategies, even though subjects were intensively briefed regarding imagination strategies prior to measurements. References: 1. Rao,Neurology(1993) 2. Schnitzler,NeuroImage(1997) 3. Nair,CogBrainRes(2003) Acknowledgements: This work has been supported by the Austrian National Bank (ÖNB-JF9305) and the Hochschuljubiläumsstiftung of the City of Vienna(1472/2002).

S142

Functional MRI: Applications

258 Reduced BOLD response to periodic visual stimulation L. M. Parkes, C. Kerskens, D. G. Norris; F.C. Donders Centre for Cognitive Neuroimaging, Katholieke Universiteit Nijmegen, Nijmegen, NETHERLANDS. Introduction: Repetitive stimulation by a periodically flickering light causes visual cortex neurons to become entrained. After several hundred milliseconds the neurons synchronise their firing to the frequency of the flickering light. In this work we investigate the BOLD response to entrainment, induced by a periodically flashing checkerboard. This is compared to random bursts of neuronal firing induced by a checkerboard flashing aperiodically, with the same average number of flashes per unit time. The BOLD amplitude change is measured for both stimuli at a range of frequencies (4 – 20Hz). Methods: Six volunteers (mean age 29yrs, range 21–42, 1F ) were recruited. The stimulus (see Figure 1) consisted of 60s of fixation cross on a black background followed by 20s of flashing checkerboard pattern. Periodic and aperiodic stimuli were presented for a range of frequencies (4–20Hz). All experiments were performed using a Siemens 3T Trio scanner. Images were acquired using a gradient-echo EPI sequence (TR 1s, TE 40ms, Bandwidth 1860kHz, flip angle 65°) with 10 slices (22.4cm FOV, 64 by 64 matrix, 3.5mm thickness, giving isotropic voxels of 3.5mm3) positioned through the visual cortex. A 3D MPRAGE sequence with 1mm3 isotropic resolution was used for the anatomical scan. Activated regions were identified and the average percentage BOLD increase for each condition and frequency were recorded for all subjects. The normalised BOLD increase for each subject was then averaged over all six subjects for each stimulation condition. Results: Figure 2 shows the BOLD signal change averaged over all 6 subjects for each condition with increasing frequency. Between 10 and 15Hz the periodic stimulus gives a reduced BOLD signal change compared to the periodic stimulus. Discussion: Based on a linear model for the addition of haemodynamic responses to repeated stimuli, the responses would be expected to be the same for both aperiodic and periodic stimuli, since they contain the same number of flashes with the same mean gap length. Additional data using a range of aperiodic conditions (not shown here) supports this view. This indicates that reduction in metabolic demand for the periodic stimulus is due to the oscillatory nature of the stimulus, causing entrainment of the visual cortex neurons. We suggest that because of the selective frequency band (10–15Hz) of this effect, it could be connected to synchronised reverberations around an internal feedback loop, involving a reduced network of active neurons.

259 Functional MRI of the visual system of the rat upon stroboscopic light stimulation N. Van Camp1, M. Verhoye1, C. C. I. De Zeeuw2, A. Van der Linden1; 1Bio-Imaging Lab, University of Antwerp, Antwerpen, BELGIUM, 2Department of Neuroscience, Erasmus Universiteit, Rotterdam, NETHERLANDS. Introduction: Different visual brain structures and nuclei respond differently to light stimuli of various frequenciesi,ii. It was our aim to investigate this frequency dependent activation in rats using fMRI and focussing on brain stem (the inferior olive and the superior colliculus) and cerebellar (nodulus and flocculus-paraflocculus) nuclei and structures. Methods: The fMRI study was performed on male Long-Evans rats (n=3,350g), anaesthetized with domitoriii (induction: 0.25mg/kg,maintenance 0.1mg/kg,IM). MRI was acquired on a SMIS MR microscope (MRRS) with a 7 T horizontal bore magnet and 8cm aperture self-shielded gradients with a strength of 0.1 T/m (Oxford Instruments), using a surface receiving- and helmholtz transmitting-coil. For the fMRI, a T2*-weighted multislice GEsequence was used (FOV:35mm, slice-thickness: 1mm, acquisition-matrix: 128x64, TR/TE: 300/14msec). Using a strobe unit positioned in front of the magnet, both eyes of the rat were stimu-

Functional MRI: Applications lated simultaneously with different frequencies of light (1-8-10-1220-50Hz). The data were processed with Medx software (Version3.41, Sensor Systems,Inc,Sterling,USA). For each structure and applied light frequency the number of pixels with a significant z-score (p<0.05) was counted and analyzed with a non-parametric Kruskal-Wallis test. Results: Concerning the brainstem, stroboscopic light of different frequencies consistently activated the superior colliculus throughout all experiments and in all rats. The size of the activated area was significantly different for the different frequencies tested (p=0.04). 1Hz and 20-50Hz seemed to provide the smallest activation area, while 8, 10 and 12Hz provided the largest areas. The inferior olive and cerebellar structures also displayed activations, but not consistently in all animals and no significant differences could be observed between different brain structures upon stimulation with different light frequencies. Discussion: We observed the highest BOLD activity in the SC, at light frequencies between 8 and 12Hz. This confirms electrophysiological and c-fos studies on this structure, which have shown that SC reacts optimal to this frequency rangei,iv. Secondly we demonstrated that light flickering above 20Hz represents the rat’s threshold above which the difference between scattering and continuous light cannot be discerned anymore (flicker diffusion). It seems that only a small part of the inferior olive (dorsal cap of Koojii) is influenced by stroboscopic light. This probably explains the lack of observing any frequency dependent activation since this structure is far too small to be distinguished upon BOLD contrast changes. The same may be true for the cerebellar structures. Endnotes: i Price,N.S.C. et al.(2001)J Neurophysiol85:1512–1521 ii Buchtel HA et al.(1972)Exp Brain Res15(3):278-88. iii Sommers MG(2002)ISMRM,Honolulu iv Correa-Lacarcel J et al. (2000)J Chem Neuroanat Mar;18(3):135-46

260 Fingersomatotopy in area 3b of the sensory cortex assessed with fMRI at 3T D. van Westen1, P. Fransson2, J. Olsrud1, G. Lundborg3, B. Rosén3, E. Larsson1; 1Dept of Radiology, University Hospital, Lund, SWEDEN, 2Dept of Clin Neuroscience, Karolinska Institutet, Stockholm, SWEDEN, 3Dept of Hand Surgery, University Hospital, Malmö, SWEDEN. Introduction: The primary sensory cortex (S1) consists of four body surface maps, areas 3a, 3b, 1 and 2 of which 3b is considered to be 'most' area S1 [1]. Fingersomatotopy in area 3b has been demonstrated with fMRI at 1.5 T [2]. The purpose of this study was to assess if spatial discrimination in the hand area could be improved at 3T. Two types of data analysis were compared. Subjects and Methods: In twenty volunteers fMRI was performed on a 3T head scanner (Siemens Allegra) using T2*-weighted SSEPI 64 x 64 matrix, 49 slices, 3 mm3 voxels, TE/TR = 30/3000ms. Sensory stimulation consisted of gentle brushing of fingers 1-5 of the right hand sequentially in a blocked paradigm with 4 'ON' periods of 10 images. Images were resampled to 1.5 mm3 voxels before motion correction, coregistration with anatomical data and smoothing with a 4 mm FWHM kernel were performed in SPM99. Stimulation effects were compared to rest for each finger in each individual, threshold p <0.05 (corr, 'MAIN EFFECT'). Secondly, each finger's activation in area 3b was contrasted with activation of the other fingers in that area; mean peak activation was thresholded

S143

at p<0.001 (unc, 'DIFFERENTIAL EFFECT'). Coordinates in [x, y, z] for the peak activation in area 3b, defined as the posterior wall of the central sulcus, were determined. Peak activation for finger 1 was set at [0, 0, 0] and distances in each direction [x, y, z] to fingers 2-5 were calculated. Results: Data from 5 volunteers were excluded due to major motion artefacts (1), global lack of activation (1), and lack of activation in area 3 b for one or more fingers (3). Mean distance [mm] for finger 2-5 to fingers 1 for 'MAIN' and 'DIFFERENTIAL' in the remaining fifteen volunteers, x-, y- and z-coordinates as well as the Euclidian distance are presented in the table. ‘MAIN EFFECT’

‘DIFFERENTIAL EFFECT’

Finger X

Y

Z

Euclidian

X

Y

Z

Eucl.

1

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

0 (0)

2

3.8 (1.7)

0.0 (1.8)

5.7 (1.1)

11.4 (1.7)

3.4 (1.2)

0.0 (1.2)

9.8 (1.1)

9.8 (1.0)

3

4.2 (1.6)

1.1 (1.5)

7.3 (1.3)

11.9 (1.6)

4.9 (1.0)

1.1 (1.2)

7.3 (1.4)

12.0 (1.4)

4

8.1 (1.7)

4.6 (1.7)

10.2 (1.6) 16.6 (1.9)

6.5 (1.3)

4.6 (1.1)

10.2 (1.4) 15.8 (1.5)

5

8.7 (2.0)

7.3 (1.2)

11.1 (2.1) 18.6 (2.2)

7.9 (1.0)

7.3 (1.0)

11.1 (1.1) 17.6 (1.0)

Conclusion: Fingersomatotopy in area 3b in S1 was demonstrated at 3 T for tactile stimulation and analysis on an individual level. Stimulation effects compared to rest ('MAIN')and each finger´s activation contrasted with activation of the other fingers ('DIFFERENTIAL') gave similar results. The Euclidian mean distance between fingers 1 and 5 of 18.6 mm ('MAIN') and 17.6 mm ('DIFFERENTIAL') was similar to results from a previous study [2]. References: 1. Kaas JH [1983] Physiol Rev 63: 206-231 2. Kurth R et al [2000] NeuroReport 7:1487-1491

261 Optimised 3T EPI enables fMRI of amygdala activation during emotion processing and induction S. D. Robinson1, G. Sachs2, W. Gombas2, C. Windischberger1, A. Rauscher1, M. Barth3, E. Moser1,3; 1Arbeitsgruppe NMR, Institut für Medizinische Physik, Universität Wien, Vienna, AUSTRIA, 2Department of Psychiatry, University and General Hospital, Vienna, AUSTRIA, 3Department of Radiodiagnostics, University and General Hospital, Vienna, AUSTRIA. Introduction: Boundaries of volumes of differing magnetic susceptibilities lead to grave signal loss and geometric distortion in low resolution EPI of inferior regions of the brain. This has given rise to doubts of the veracity of the many recent claims of amygdala activation (Merbolt et al., 2001). We have optimised BOLD-EPI at 3T via a systematic study of the most suitable slice orientation, resolution, slice thickness and effective echo time. The resultant protocol has been found effective in charting amygdala activation in two fMRI studies of emotion processing and experience. Subjects and Methods: All measurements were carried out using a wide-bore 3 T Bruker Medspec S300 (Bruker-biospin, Ettlingen, Germany), a BG-A55 gradient coil and a standard head coil. For EPI protocol optimisation, ten subjects were examined with singleshot EPI protocols, with matrix sizes varying from 64x64 to 256x256, a 25x21 cm FoV, slice thicknesses ranging from 1mm to 6mm, and receiver bandwidths (RBWs) between 100 and 200 kHz. Optimum parameters were established via SNR calculation in the amygdalae. All fMRI measurements were performed with the optimised GE-EPI protocol (oblique 2mm axial slices, 128x128 matrix, TEeff=46 ms, with 12 axial slices in TR=2s). Stimuli were the emotion discrimination task of Gur et al (2002) used with 14 schizophrenic patients and 17 controls, and a number series task in

S144

Functional MRI: Applications

which items of increasing difficulty induced exasperation and learned helplessness in a proportion of healthy subjects. Results:

(Bach) and white noise, all having the same 90 dB sound level. The stimulation period lasted for 30 seconds following a rest period of the same duration. In each period 6 images were acquired. This block 'off' vs. 'on' was repeated 30 times. Results: Field L, the bird's analogue of the mammalian auditory cortex, was activated upon each of the applied stimuli, while NCM (caudal medial neostriatum), the region considered as substrate for auditory recognition memory, was exclusively activated upon hearing birdsong (see arrows in figure). Besides the detection of activated regions, it is also possible to measure the temporal aspects of neuronal activity. On the long term (30 min) habituation was not observed, but within a well-defined short period (30 sec) habituation occurred in both brain regions and with each stimulus. In field L the auditory response upon white noise reached its maximum immediately after onset of stimulation, whereas with conspecific song a maximal response was reached after 10-15 seconds.

T2* in the amygdalae was measured to be 46±2 ms, compared with 51±1 ms in the temporal cortex (Klarhoefer et al., 2002). A matrix size of 128x128 was found to be satisfactory to reduce signal loss so that the amygdalae could be delineated. Image distortion was minimised using a RBW of 200 kHz. In the amygdalae, SNR was found to peak at a slice thickness in the region 2.0 - 2.5 mm, as shown in Figure 1. FMRI analysis indicates activation in the amygdala of 13/15 subjects under the emotion discrimination task (Figure 2) and a 'highly reactive' subject studied using the number series task. Conclusion: The high resolution EPI protocol described has proved capable of detecting robust activation in the amygdala. References: 1. Merboldt KD et al. (2001). NeuroIm 14: 253-257. 2. Gur RC et al. (2002). NeuroIm. 16: 651-62. 3. Klarhöfer M et al. (2002). Mag.Res.Im. 20: 359-364

262 Functional Magnetic Resonance Imaging (fMRI) of the brain of songbirds: a new tool for the cognitive neuroscience of auditory learning and memory T. Boumans, V. Van Meir, M. Verhoye, J. Van Audekerke, A. Van der Linden; Bio-Imaging Lab, University of Antwerp, Antwerp, BELGIUM. Introduction: The species-specific vocalizations of songbirds are crucial for their social interactions and reproductive success. Like in humans, a learning process during childhood is necessary to produce socially relevant songs in adulthood. The brain circuitry involved in hearing, learning and producing songs is very well documented in songbirds, and make them an excellent model for studying the activity of brain regions during cognitive tasks and learning processes. In this study we used BOLD fMRI to investigate brain activity upon repetitive exposure to different songs and sounds within the same bird. Subjects and Methods: Five European starlings were anesthetized (IM, chest) with a ketalar-domitor mixture and immobilized in a non-magnetic stereotaxic head holder combined with a radio-frequency surface antenna (24 mm) and HelmHoltz (50 mm). Imaging was carried out on a 7T MR microscope with horizontal bore (SMIS, UK). Bilateral, a sagittal slice was imaged, 0,9 mm from the midline. A T2*-weighted multi-slice gradient-echo sequence was used with TR 80 ms, TE 14 ms, acquisition matrix 128x64, FOV 30 mm, slice thickness 0,8 mm. Functional MRI data were obtained from anesthetized birds upon listening consecutively to unfamiliar conspecific birdsong (other starlings), complex song

Discussion: Neuronal plasticity constitutes the basic mechanism on which perception, memory and cognition in general are based. Auditory fMRI in songbirds, using rapid image acquisition, will allow us to study plastic neuronal properties underlying song learning and auditory memory in a well documented animal model.

263 Dynamic Manganese Enhanced (DME) MRI reveals the song selective response of specific neuronal populations in the song control system of the canary I. Tindemans1, M. Verhoye1, J. Balthazart2, A. Van der Linden1; 1Bio-imaging Lab, University of Antwerp, Antwerp, BELGIUM, 2Center for Cellular and Molecular Neurobiology, University of Liège, Liège, BELGIUM. Introduction: One of the best studied song control nuclei in the songbird brain is HVC (formerly known as High Vocal Center) which contains interneurons, neurons projecting to area X and neurons projecting to the nucleus Robustus Archistriatalis (RA) ( Fig. a:schedule, b:MRI-image).

Animal models: Spectroscopy

These 3 neuron types display specific differences and show songselective responses to auditory stimuli. Using in vivo DME-MRI1 after MnCl2 injections into HVC we investigated which type of HVC neurons are activated upon exposure to conspecific song (COS). Methods: A non-metallic 22G cannula (Plastics One Inc., USA) was implanted in the HVC of 7 adult male canaries at least 2 weeks before the first MRI experiment. The anaesthetized (0.8% isoflurane, 7:3 N2O:O2) and accurately monitored (temperature, pCO2, respiration rate) canaries were immobilized in a customized stereotaxic device with an integrated RF surface receiver (Φ_upper;=17mm) and Helmholtz transmit (Φ_upper;=45mm) antenna. Inside the magnet, 200nl 0.01M MnCl2 was injected into HVC through the permanent cannula. Subsequently coronal images were obtained during 8 hours with a T1-weighted multislice SE sequence: 265x128 matrix, TR/TE=300ms/10ms, flip angle=90°, 25mm FOV, 10 slices of 0.8mm thickness, 16 averages (7T MRRS system, U.K.). This procedure was repeated twice (14 days in between) in the same bird with (COS+) and without (COS-). Results: The relative changes in time of the manganese enhanced signal intensity in RA and area X were fitted to sigmoid curves which defined the SImax, and the ‘n coefficient (describing the curve shape) reflecting the uptake kinetics of Mn2+ in both COS+ and COS- experiments. Wilcoxon tests revealed that the n parameter was significantly higher in RA (p<0.05) for COS+ experiments, while a significant difference in SI max (p<0.05) was observed in area X between COS- and COS+. Discussion: Hearing conspecific song induced a significant change in the shape of the Mn uptake curves in HVC neurons projecting to RA and in the amount of Mn transported from HVC to Area X. This indicates that hearing COS affected differentially both classes of HVC projection neurons. DME-MRI thus allows assessing the functional state of specific neuronal populations in the song system of living canaries in a manner reminiscent of fMRI (but with higher resolution) or of 2-deoxyglucose autoradiography (but in living subjects). 1. Van der Linden A, Verhoye M, Van Meir V, Tindemans I, Eens M, Absil P, Balthazart J [2002] Neurosci 112(2):467-474

Scientific Session Animal models: Spectroscopy 4:30 pm - 6:10 pm

Jurriaanse

264 In vivo CSI of glutamate in Macaca mulatta brain C. Juchem1, H. Merkle2, N. K. Logothetis1, J. Pfeuffer1; 1MR Imaging and Spectroscopy, Max Planck Institut für biologische Kybernetik, Tübingen, GERMANY, 2Laboratory of Functional and Molecular Imaging, NIH/NINDS, Bethesda, MD. Introduction: Glutamate is the main excitatory neurotransmitter in

S145

the CNS and plays together with glutamine an important role in brain physiology. A quantitative spatially resolved analysis of glutamate separate from glutamine is therefore of particular neuroscientific interest. In a pilot 1H Chemical-Shift Imaging (CSI) study we demonstrate that maps of sufficiently high spectral and spatial resolution can be obtained to initially map areas of largely different glutamate concentration (e.g. neural tissue vs. ventricles). Methods: Measurements were made on a vertical 7T/60cm Bruker Biospec system as described previously [1]. A surface coil of 80mm diameter [2] was used for anatomical scout images shown in figure 1 (MSME, TE=100ms) and for the CSI study.

Shimming with FASTMAP led to a water line width of 13Hz in the selected 28x28x4mm3 axial slice through the ventricles. For CSI, a STEAM sequence was used with an 8x8 phase encoding scheme, leading to a nominal in-plane resolution of 3.5x3.5mm2 (TE=TM=10ms, TR=4s, NA=35). Water suppression was achieved by a VAPOR module, consisting of 7 CHESS pulses. Mild Gaussian apodization (exp(-0.5) at the edge) and spatial zero filling to 16x16 was applied before Fourier transformation. Quantification in the frequency domain was done voxelwise with LCModel relative to total creatine assuming 10mM in brain matter. Results: Figure 2 shows a typical spectrum of brain tissue, demonstrating the spectral separation of the glutamate and glutamine multiplets at 2.35 and 2.45ppm, respectively. Image smoothing of a reduced FOV of 21x24.5mm2 (yellow frame in Fig.1) led to the glutamate map in Fig.3. For brain tissue Cramer-Rao bounds are in the range of 6-10% and metabolite peaks are 7-11Hz broad. The ventricle anatomy is assigned by a white contour that fits very well with the expected low glutamate concentrations within the ventricles.

S146

Animal models: Spectroscopy 265 Lithium spectroscopic imaging of rat brain at therapeutic doses S. Ramaprasad; Radiology, University of Nebraska Medical Center, Omaha, NE. Introduction: Since the discovery that lithium (Li) is efficacious for the treatment of manic-depressive illness, the brain lithium level and the distribution of lithium across the brain have been of interest. Differences in Li levels in various brain regions for autopsied rats and humans (1-5) have been reported. However, many results are conflicting due to ion redistribution at death and/or during sample preparation. The application of MR technology to rat brain regions has provided information on lithium distribution in a noninvasive manner (6, 7). These earlier images generated in the during the developmental phase used high dose of lithium. Here we present the results of our 7 tesla spectroscopic imaging studies on rat brain under therapeutic doses. Subjects and Methods: Animal preparation: Male Sprague Dawley rats weighing 200-300 g were housed under controlled lighting and temperature. Food and tap water were available ad libitum. The Li protocol consisted of two doses of LiCl on the 1st and 2nd day followed by a single dose on the morning of the third day. MR imaging studies were done 1-2 hours following the final dose. The 2meq doses lead to serum levels comparable to human therapy (1.1±0.3 meq/liter, n=7). The MR parameters used are: 1 H Imaging: 256x256 Matrix; TE, 20-30 ms; pulse repletion times of 0.5-1s, slice thickness of 5mm;Fov of 40 mm. 7 Li Imaging: 8x8 matrix; TR of 1s;Number of acquisitions 8-16;slice thickness 10mm;Fov of 40mm. Spectroscopic images are presented after zero filling once before Fourier transformation.

Discussion: In most CSI studies insufficient sensitivity or long TE values do not permit the quantification of glutamate+glutamine. Typically glutamate and glutamine can't be separated due to limited spectral dispersion/resolution. In this study, utilizing high magnetic field, we were able to separate glutamate and glutamine and to measure a pure CSI glutamate map in the primate. The results suggest that the resolution we are likely to obtain in the near future by using implanted coils may permit differentiation of brain structures in the millimeter range (gray vs. white matter) and/or detection of small concentration differences in the same structure (activated vs. non-activated cortex). 1. ISMRM 1877,2089(2003) 2. MAGMA 15,425(2002) 3. Neuron 35,227(2002)

Figure1: An overlay of Li SI data on the proton image of the rat head. The voxel volume was 63 µl. Results: The results shown in Figure 1 provide the spectroscopic image of lithium distribution in the brain and surrounding tissue at therapeutic dose. Discussion: These results demonstrate the first application of spectroscopic imaging at 7T field strength at therapeutic doses. References: 1. Ebadi MS, Simmons VJ, Hendrickson MJ, Lacy PS. Eur. J. Pharmacol. 27, 324, 1974. 2. Bond PA, Brooks BA, Judd A. Br. J. Pharmacol. 53, 235, 1975. 3. Mukherjee BP, Bailey PT, Pradhan SN. Psychopharmacology. 48, 119, 1976. 4. Ebara T, Smith DF. J. Psychiatr. Res. 15, 183, 1979. 5. Francis RI, Traill MA. Lancet ii. 523, 1970. 6. Ramaprasad et al. Mag. Reson. Med. 25, 308-318,1992.

Animal models: Spectroscopy 7. Ramaprasad S. Proc. Intl. Soc. Mag. Reson. Med. 10, 1216, 2002 Acknowledgements: This research was partly supported by NARSAD.

266 In vivo MRI and MRS study of a murine model of cerebral malaria M. Penet1, A. Viola1, D. Ibarrola1, S. Confort-Gouny1, Y. Le Fur1, N. Coltel2, M. De Reggi3, B. Gharib3, P. Cozzone1; 1Faculté de Médecine La Timone, Centre de Résonance Biologique et Médicale, UMR-CNRS 6612, Marseille, FRANCE, 2Faculté de Médecine La Timone, Unité de Parasitologie Expérimentale, Marseille, FRANCE, 3Faculté de Médecine La Timone, INSERM U 399, Immunologie et Génétique des Maladies Parasitaires, Marseille, FRANCE. Introduction: Malaria is a major tropical disease, causing more than one million deaths per year. Cerebral involvement in the course of Plasmodium falciparum infection is the most common lethal complication with a mortality rate up to 50 % and residual disability in many surviving subjects. The aetio-pathogenesis of cerebral malaria leading to an acute encephalopathy is unclear and the current hypothesis relies either on the sequestration of parasited erythrocytes in the nervous system that would impair brain microcirculation, or alterations of brain hemostasis. Here, we present the results of an MRI/MRS study of a murine model of the disease demonstrating the coexistence of structural and metabolic alterations. Subjects and Methods: In total 12 female CBA/J mice were infected with Plasmodium berghei ANKA and explored at day 6 and 7 post-infestation. In addition, 12 healthy mice of the same strain were used as controls. The explorations were performed at 4.7 T on a Bruker AVANCE Biospec spectrometer. The brain MRI protocol included multi-slice spin echo T2-weighted images and T1-weighted images. T1-weighted images were acquired before and after the injection in the caudal vein of a contrast-agent (Magnevist). Localized proton spectrometry (1H MRS) of brain was obtained by using the PRESS sequence (time of echo, 135 ms; volume of interest, 3.5x3.5x3.5 mm3; 256 scans). Results: MR images disclosed in diseased mice diffuse and symmetrical hyperintense areas on axial T2-weighted images, which were enhanced on post-contrast T1-weighted images, mostly in the peri-striatal fibers and the corpus callosum. A notable increase in brain volume was also observed. Axial T2–weighted images displayed diffuse and symmetrical hypointense signals in sub-arachnoid spaces. Brain MRS showed a statistically significant reduction in N-acetylaspartate (NAA) in infected mice (NAA/S = 0.283±0.01 (mean ± sd) (n=8) S=sum of all metabolites; healthy controls: NAA/S = 0.340 ±0.008 (n=5); P<0.005), a decrease in creatine (Cr) (infected mice: Cr/S= 0.295±0.010 (n=8), healthy controls: Cr/S = 0.319±0.018, (n=5); P<0.05) and the apparition of lactate. Discussion and Conclusion: Our results demonstrate the existence of an important brain oedema in cerebral malaria with a sub arachnoid haemorrhage and lesions associated with the rupture of the blood brain-barrier. The metabolic studies show an important alteration of the neuronal metabolism and a decline in creatine, a compound related to the cellular energetic processes whereas no change in choline-derived metabolites was observed. The presence of lactate is suggestive of an ischemic process whose origin remains to be elucidated.

S147

267 13C NMR characterization of transgenic knock-out mice in D1 dopamine receptors as models of Parkinson disease T. B. Rodrigues1, P. Sanchez1, A. Sierra1, M. Benito1, P. Ballesteros2, C. Geraldes3, A. B. Martín4, R. Moratalla4, S. Cerdan1; 1NMR Laboratory, Instituto Investigaciones Biomédicas, Madrid, SPAIN, 2Organic Synthesis and Molecular Imaging, Instituto de Investigación, Madrid, SPAIN, 3Department of Biochemistry, University of Coumbra, Coimbra, PORTUGAL, 4Plasticidad Neuronal, Instituto de Neurobiología Ramon y Cajal, Madrid, SPAIN. Introduction: Dopaminergic activity is modulated by the primary stimulatory dopamine receptor-type D1. These receptors play an important role in the mechanisms of some drug addictions (e.g. cocaine) and in Parkinson disease. Previous reports1 have shown evidence that D1 receptor stimulation reduces the extracellular glutamate concentration (Abekawa et al. Brain Res. 2000, 250-4) while other authors have reported that inhibition of the metabotropic glutamate receptor modulates D1 receptor activation (David and Abraini, Eur. J. Neurosci, 2001, 13, 2157-2164). These observations are consistent with the existence of interactions between dopaminergic and glutamatergic neurotransmissions. Carbon-13 NMR spectroscopy has been shown to be a particularly useful tool to measure glutamatergic neurotransmission and the glutamineglutamate cycle activity. Here we investigate the effect of D1 receptor suppression in knock out mice on the activity of glutamine-glutamate cycle. Methods: D1 receptor knock-out mice were generated at Instituto Cajal CSIC (Madrid) with full D1 expression (+/+ homozygotes), one half D1 expression (+/- heterocygotes) or lack of D1 expression (-/- homocygotes). All animals were anaesthetized with isoflurane, receiving a 60 minutes infusion of (1,2-13C2) acetate (24 mols.min-1.100g-1) in the left jugular vein. The brains were then funnel frozen and extracts were prepared and analyzed by 13C NMR (125.13 MHz, pH 7.2; 25ºC) with proton-decoupling only during the acquisition. 13C NMR spectra were simulated completely using the WINDAYSY, calculating the areas relative to the unchanged inositol carbons. Results: The Table summarizes the relative 13C incorporation in the doublet resonances of cerebral glutamate (GluC4d) and glutamine C4 (GlnC4d) in the different mice used in this study. Mice type GluC4d, GlnC4d D1 +/+ 2.924 ± 0.313, 3.090 ± 0.352 D1 +/- 2.527 ± 0.148, 3.923 ± 0.759 D1 -/-, 2.843 ± 0.529, 3.832 ± 1.443 D1 receptor expression did not affect apparently glutamate C4 labelling. However, it increased glutamine C4 labelling in (+,-) and (/-) mice, revealing an increase in glutamine synthase activity. Discussion: Our results confirm an important interaction between dopaminergic and glutaminergic neurotransmissions. Removal of D1 receptors, appears to induce an increase in the glutamine cycle activity and glutamatergic neurotransmission. This implies that neurological disturbances associated to drug addiction or Parkinson disease may be investigated in the human brain by 13C NMR. Moreover, our results suggest that a successful treatment these disorders must address disturbances in both, glutamatergic and dopaminergic neurotransmissions. Acknowledgements: This research was covered in part by strategic group grant from the Community of Madrid to PB.

S148

Animal models: Spectroscopy

268

269

2D J-resolved Spiral Spectroscopic Imaging on rat brain tumour B. Hiba, R. Serduc, P. Provent, R. Farion, C. Rémy, A. Ziegler; Neuroimagerie Fonctionnelle et Métabolique, INSERM U594, Grenoble, FRANCE.

Membrane phosphatidylethanolamine as source for 1H NMRvisible polyunsaturated lipids in vivo J. M. Hakumäki, T. Liimatainen, S. Ylä-Herttuala; A.I.Virtanen Institute, University of Kuopio, Kuopio, FINLAND.

Purpose: Lactate and lipids are often detected in the 1H MR spectra of brain tumours. As both signals overlapped in the 1.3 ppm region of the spectra, it is difficult to determine the precise distribution of each species when analysing spectroscopic images. The 2D J-resolved spectroscopy may be used to unambiguously separate the two resonances: lactate will appear as a doublet in the f1 domain, and the lipid signal will appear as a singlet. To localize these metabolites with a high resolution within a reasonable duration, a rapid spiral spectroscopic imaging sequence was used. The method was applied to map the lactate and the lipid detected in C6 rat brain tumours at late growth stage. Subjects and Methods: Intracerebral tumours were induced by injection of C6 cells in 5 rats (Wistar, male, 220 g). NMR was performed at 7 T, 24 days after implantation. Rats were anaesthetized with a mixture of isoflurane in air and 30% O2, and fixed with ear bars to a home-built probe. The sequence consisted in a water suppression module, an outervolume saturation module, followed by a PRESS sequence and spiral readout gradients. The echo-time was incremented from 116 ms to 636 ms, by step of 26 ms to describe the t1 domain. The spiral readout module consisted in 8 spirals of 8 turns each (total duration 81.92 ms); 32 temporal interleaves were used to describe the t2 domain (bandwidth = 3125 Hz). With 2 spatial interleaves, the FOV was 30 mm. The voxel volume was 8 mm3. Data are processed according to four steps: 1) data reordering, 2) first-order phase correction to project data onto regularly spaced planes in t2 domain, 3) spatial resampling onto a spatial Cartesian grid, 4) Hanning filtering, zero filling and FFT in the 4 domains. Results: Localized 2D J-resolved spectra were obtained throughout the FOV (Fig.1). Total acquisition duration was 64 min. By integrating the peak volume in each voxel, maps of lactate and lipids were reconstructed (Fig. 2). The lipid signal was localized mainly in the center of the tumour corresponding mainly to necrosis. Lactate signal was detected everywhere in the tumour. Discussion/Conclusion: 2D-spatial 2D-spectral spectroscopy was performed on small animals with a high spatial resolution in about 1 hour. Mobile lipids were found mainly in the necrotic part of the tumour. Lactate was produced both by viable and hypoxic tumour tissue.

Purpose/Introduction: 1H NMR visible lipids can be observed in pathological conditions, such as cancer. Recent evidence supports the notion that the signals originate from intracellular lipid bodies (1). We have recently demonstrated, that polyunsaturated fatty acids (PUFA) accumulate in tumors undergoing apoptosis in vivo (2). Here, 1H 2D COSY NMR spectroscopy of excised tumors, transmission electron microscopy (TEM) and a biochemical analysis of main membrane constituents (75-80% of total lipid), phophatidylcholine (PtdCho), phosphatidylethanolamine (PtdEth) and lysophosphatidylcholine (lysoPtdCho) have been performed to study membrane involvement in this accumulation process. Subjects and Methods: Intracranial HSV-tk positive BT4C gliomas were inoculated in BDIX rats,and treated with ganciclovir (GCV) for 8 days as described before (2). Animals were sacrificed prior to, and on days 4 and 8 of treatment (n=4). Standard 2D COSY spectra were acquired at 9.4T from excised brain tumors (n=4) in 0.9% NaCl D2O ex vivo (4-5 Hz shim lw, F1=512 and F2=512 after zero-filling, SW=5 kHz). Individual phospholipids from excised tumors (n=4-5) were separated by thin-layer chromatography and fatty acid components were analyzed by gas liquid chromatography. Standard TEM sections (n=3-4) were quantified for intracellular lipid bodies. Results: In 2D COSY spectra of untreated tumors, lipid cross peaks from the saturated methylene and methyl resonances could be observed. However, after 4 and 8 days of GCV treatment a strong cross peak corresponding to polyunsaturated bis-allylic fatty acyl chains could be observed. TEM analysis of lipid bodies corresponded with a 38% increase in droplet volume. Chromatographic analysis of tumor lipids showed significant changes in the membrane PtdEth fractions of monounsaturated (reduced by 49%) and polyunsaturated fatty acids (reduced by 73%). No significant changes in PtdCho or lysoPtdCho could be observed.

Discussion/Conclusion: The primary component of 1H NMR visible PUFA resonances in BT4C gliomas undergoing apoptosis appears to be polyunsaturated bis-allylic fatty acyl moieties stored into lipid bodies. This is supported 2D NMR, and biochemical analyses showing that membrane PtdEth lose a significant fraction of these moieties (PUFA) during treatment with GCV. Such

Animal models: Spectroscopy

S149

membrane alterations strongly support phospholipase A2-activity in tumor cells undergoing apoptosis, since this enzyme cleaves fatty acyl moieties from the sn-2 positions of PtdEth, which are richly polyunsaturated. Our results suggest a rational approach for assessing actual phospholipase activity by 1H NMR spectroscopy and/or spectroscopic imaging of tumor lipids. 1. Hakumäki JM and Kauppinen R.A. (2000) TIBS 25:357-362. 2. Hakumäki JM et al. (1999) Nature Med. 5:1323-1327.

270 In Vivo 19F MR studies of fluorine labeled photosensitizer in murine tumor model S. Ramaprasad1, E. Rzepka1, S. S. Joshi2, M. P. Dobhal3, J. Missert3, R. K. Pandey3; 1Department of Radiology, University of Nebraska Medical Center, Omaha, NE, 2Departments of Genetics, Cell Biology, and Anatomy, University of Nebraska Medical Center, Omaha, NE, 3Photodynamic Therapy Center, Roswell Park Cancer Institute, Buffalo, NY. Introduction: Photodynamic therapy (PDT) is a cancer treatment modality that combines light sensitive drug and lasers [1]. Monitoring the photosensitizer (PS) in the tumor and in normal tissue is helpful in the development of new photosensitizers. Syntheses of labeled photosensitizer that can be monitored by MR studies offer the advantage of noninvasive assessment of the photosensitizer in a single subject [2]. The assessment of the same in the skin and underlying muscle may provide information about the cutaneous toxicity of PS. In this work we present the construction of pharmacokinetic profile of a new photosensitizer in a tumor model and its utility in PDT studies. Methods: The Radiation induced fibrosarcoma (RIF) cells were maintained according to the protocol of Twentyman et al [3]. Tumors were grown on mouse foot dorsum by inoculating 2x105 cells. The photosensitizer was administered IP ( ~100µM). 19F MR spectra were collected on a Bruker 7T instrument using a home built surface coil. The 19F MR spectral parameters included a 90° pulse of 16µs, a spectral width of 20 KHz, 8K data points, and a 2s repetition time for a total accumulation time of 30 minutes. PDT measurements were performed at 630 nm with an argon ion (Spectra physics model 2045) pumped dye laser (Spectra Physics, 375B). Laser irradiation was done for 30 minutes at a power of 150 mW cm-2 leading to a total light dose of 270 J cm-2. Results: In this presentation we report the results obtained using a newly synthesized water soluble form of fluorine labled photosensitizer which was monitored in mouse tumor model over time. The measured relaxation times in the solution were 924±38 ms for T1 and 150±2 ms for T2 and were used to optimize of tumor spectra. The Pharmaco- kinetic profile of PS in tumor model was constructed using three tumors (Figure 1). Based on this profile PDT was performed at time points 2hr, 4hr and 24 hrs post drug administration. Conclusions: MR studies can provide quantitative data on photosensitizer in tumor and a rational basis for PDT initiation. PDT studies done at 2hrs post PS administration led to tumor regression between days 2-4. References: 1. Dougherty, TJ, et al. [1998] J.Natl. Cancer. Inst. 41:351-359. 2. Ramaprasad et al [1994] Proc. Soc.Magn. Reson Med. 3: 1348. 3. Twentyman et al [1980] J.Natl. Cancer. Inst.64: 595-604. Acknowledgement: This research was funded by the Department of Defense (DAMD 17-99-1-9065).

271 Effects of post-ischemic NHE blockade on [Na+]i and pHi in rat hearts M. Ten Hove1, C. J. A. Van Echteld2; 1Experimental Cardiology, Interuniversity Cardiology Institute of the Netherlands, Utrecht, NETHERLANDS, 2Experimental Cardiology, NMR laboratory, University Medical Center Utrecht, Utrecht, NETHERLANDS. Introduction: Ischemic blockade of the Na+/H+ exchanger (NHE) has been found to be cardioprotective in many studies. However, reports on the efficacy of specific NHE blockers when administrated only during reperfusion are inconsistent. Differences in the severity of ischemia and in drug delivery may explain these inconsistencies. Little is known about the primary goal of post-ischemic NHE blockade, i.e. reduction of Na+i overload. Methods: Isolated rat hearts were subjected to either 25 minutes of zero flow ischemia or 60 minutes of low flow (0.2 ml/min.) ischemia. Hearts were reperfused with or without the selective NHE blocker cariporide added to the perfusate. In the low flow group cariporide was administered 5 minutes prior to reperfusion. [Na+]i and pHi were measured with simultaneous 23Na and 31P MRS, respectively, on a 3 channel Bruker Avance NMR spectrometer equipped with a 9.4 T magnet. TmDOTP5- was used as a shift reagent to discriminate between Na+i and Na+e. Results: Administration of cariporide after zero flow ischemia caused the [Na+]i to decrease during the first minute of reperfusion, followed by a partial and transient rise during the second minute (fig. 1). Untreated hearts showed a very small rise in [Na+]i during the first minute. The pHi recovered slightly slower in cariporide treated hearts than in untreated hearts. No effect of cariporide on the rate pressure product could be detected since the rate pressure product recovered fully in untreated hearts. The end diastolic pressure, however, was increased during reperfusion in both groups, indicating that cariporide was not cardioprotective. After low flow ischemia the [Na+]i (fig. 2) and the pHi recovered similarly in both groups. Recovery of the rate pressure product was poorly in both groups. Conclusion: Post-ischemic NHE blockade only has a small and transient effect on [Na+]i and pHi after zero flow ischemia and no effect at all after low flow ischemia, which explains the lack of cardioprotection under these experimental conditions.

S150

Flow Quantification: Motion and Artefacts The computed mean pulmonary arterial pressure value (PpaComp) was compared with the mean pressure value obtained from catheterization (PpaCat) in each patient. Results: A significant correlation was established which did not differ from the identity line PpaComp = PpaCat (r = 0.92). The maximal and mean absolute differences between PpaComp and PpaCat were 11.9 mmHg, and 5.4 mmHg, respectively. Conclusion: We conclude that this computed method that combines biophysical (whatever the patient) and patient-wide parameters improves the accuracy of MRI to non invasively estimate mean pulmonary arterial pressure.

273 Figure 1: [Na+]i during reperfusion with or without post-ischemic NHE blockade after 25 minutes of zero flow ischemia.

Figure 2: [Na+]i during reperfusion with or without post-ischemic NHE blockade after 60 minutes of low flow ischemia.

Scientific Session Flow Quantification: Motion and Artefacts 4:30 pm - 6:10 pm

Ruys

272 A computed method for non invasive MRI assessment of pulmonary arterial hypertension E. Laffon1, C. Vallet2, V. Bernard3, M. Montaudon4, D. Ducassou1, F. Laurent4, R. Marthan5; 1Service de Médecine Nucléaire, Hôpital Haut-Lévèque, Pessac, FRANCE, 2Enseirb, avenue du Dr Schweitzer, Talence, FRANCE, 3Service de Cardiologie, Hôpital Haut-Lévèque, Pessac, FRANCE, 4Service de Radiologie, Hôpital Haut-Lévèque, Pessac, FRANCE, 5Lpcr inserm emi 0356, Université de Bordeaux 2, Bordeaux, FRANCE. Purpose: A method is described that enables to non invasively assess mean pulmonary arterial pressure from M.R. phase mapping by computing both biophysical and patient-wide parameters. Subjects and Methods: The biophysical parameters are the mean blood flow velocity over the cross-sectional area (CSA) of the main pulmonary artery (MPA) at the systolic peak, and the maximal systolic MPA CSA value. The patient-wide parameters are the height, weight, and cardiac frequency. These parameters have been measured in 31 patients undergoing right-side heart catheterization.

Hemodynamic evaluation of renal artery stenosis in renovascular hypertension: Retrospective cardiac-gated phase-contrast MR vs Doppler Flowmetry Ö. Özsarlak, P. Van Dyck, P. M. Parizel, A. M. De Schepper; Department of Radiology, University Hospital Antwerp, Antwerp, BELGIUM. Purpose: The aim of this study was to define a correlation between the degree of stenosis and peak systolic velocity changes in the renal arteries measured by 3D MR cine phase-contrast flowmetric study. The second aim was to compare the MR PC-flowmetry with the Doppler flowmetry of renal arteries in the evaluation of renovascular Methods and Materials: Thirty-five patients clinically suspected of having a renovascular hypertension underwent both MR and Doppler examinations for assessment of renal artery stenosis. MR flowmetry is obtained by using retrospective cardiac-gated 3D MR cine phase-contrast technique, after morphologic assessment of renal arteries by contrast enhanced 3D MR angiography. Two acquisitions were performed on each case through the proximal portion of renal arteries with a velocity-encoding 200 cm/s. Doppler flowmetry was also performed by calculating the peak systolic velocities through the proximal portion of renal arteries on each side. The results were compared. Results: There is a good correlation of peak systolic velocity measurements between both modalities. The peak systolic velocity threshold value of 80 cm/s detected unilateral high-grade stenosis in 6 patients and bilateral high-grade stenosis in two patients. In two patients with subocclusive renal artery disease, no flow value was able to obtain on both modalities. Conclusion: The combination of MR angiography and MR flowmetry allows both morphologic and functional assessment of the renal arteries. There is a good agreement between the MR cine phase-contrast flowmetry and Doppler flowmetry results.

Flow Quantification: Motion and Artefacts 274 MRI quantification of the role of the reflected pressure wave on coronary and ascending aortic blood flow E. Laffon1, C. Galy-Lacour1, F. Laurent2, D. Ducassou1, R. Marthan3; 1Service de Médecine Nucléaire, Hôpital HautLévèque, Pessac, FRANCE, 2Service de Radiologie, Hôpital Haut-Lévèque, Pessac, FRANCE, 3Lpcr-emi 0356, Université de Bordeaux 2, Bordeaux, FRANCE. Purpose: The respective roles of the windkessel effect, which is related to the vessel compliance, and of the returning pressure wave, which is reflected off peripheral resistances, in generating the total diastolic blood flow in the ascending aorta (VT: coronary + flow through the upper part of the ascending aorta) were assessed under physiological conditions. Subjects and Methods: In ten healthy young men (21 - 30 years), magnetic resonance phase mapping was used to assess non-invasively both blood flow and vessel cross-sectional area (CSA) at the upper part of the ascending aorta. Measurements of blood flow velocities and CSAs were carried out over complete cardiac cycles, with a 30 ms resolution time. Results: The total diastolic blood flow volume (VT) was significantly greater than the diastolic blood flow volume related to the windkessel effect (Vw). Consequently, since the windkessel effect could not account for the whole diastolic blood flow volume, the difference was attributed to the diastolic blood flow volume related to the reflected pressure wave (Vpw). Mean and SD of Vw and Vpw were estimated to be 4.34 ± 1.94 ml cycle-1 and 6.80 ± 2.40 ml cycle-1 (n = 10), respectively, for a measurement slice placed at 5.12 cm (on average) from the aortic valve and a mean coronary diastolic blood flow volume assumed to be equal to 4 ml cardiac cycle-1 in all subjects. Conclusion: The present non-invasive MR phase mapping study indicates that the windkessel effect and the reflected pressure wave might play a similar role in generating the total diastolic blood flow in the ascending aorta, under physiological conditions.

275 Retrospective respiratory motion compensation for cine coronary flow measurements using navigator echoes C. Baltes, S. Kozerke, P. Boesiger; Institute for Biomedical Engineering, University & ETH Zurich, Zurich, SWITZERLAND. Introduction: In cardiac magnetic resonance (MR) imaging navigator gating is commonly applied to compensate for respiratory motion. While this technique provides good motion compensation for short acquisition windows within the cardiac cycle, simple gating might not be appropriate in applications acquiring data throughout the cardiac cycle such as in coronary flow quantification [1]. The importance of minimal temporal delays between the navigator and actual data acquisition has been pointed out recently [2]. Due to the required high temporal resolution in coronary flow quantification, navigator echoes cannot be played out prior to each cardiac phase being sampled. The objective of the current work was to improve image quality of cardiac phases acquired later in the cardiac cycle by using retrospective image corrections based on navigator information additionally collected at the end of each cardiac cycle. Methods: A standard TFE phase contrast sequence with a single leading navigator (L-NAV) for gating and slice position correction was extended with a second trailing navigator (T-NAV) (Fig. 1).

S151

In-vivo blood flow quantification in the left anterior descending (LAD) coronary artery was performed on a 1.5 T Intera MR system (Philips, Best, The Netherlands) with the following imaging parameters: 29 heart phases, dt=37.6ms, venc=30cm/sec, resolution:0.9mmx0.9mm, angulation:24°/-19°/0°(AP/FH/RL), tracking factor:0.6 [3], gating window:5mm. During scanning, the respiratory positions detected by the navigators were recorded into a file along with the sampled k-space data. In post-processing, cardiac positions for each heart phase were computed by linear interpolation of the diaphragm positions provided by the leading and trailing navigators. This information was then used to correct the image data for in-plane components of residual respiratory motion according to the Fourier shift theorem.

Results: Figure 2 compares the acquired anatomical images and phase maps showing the LAD (arrows) before (2a,2b,2e,2f) and after (2c,2d,2g,2h) correction of respiratory motion for two different heart phases. Image blurring observed in later heart phases due to increasing delays between acquisition and leading navigator was significantly reduced (2e,2g). In addition, the correction leads to improved phase maps resulting in modified mean velocity profile and volume flow (Fig. 3).

S152

Flow Quantification: Motion and Artefacts

Discussion: In this study, a post-processing approach has been presented to correct in-plane components of residual respiratory motion in navigated cine acquisitions. The current method achieved considerably improved image quality and velocity maps in particular for phases acquired late in the cardiac cycle. References: 1. Nagel E,MRM;41,544-549,1999 2. Spuentrup E,MRM;47,196-201,2002 3. Wang Y,MRM;33,713-719,1995

276 Non-ECG-Triggered, High-Resolution Balanced Radial CMRA with affine motion compensation I. Graesslin1, M. Mittelbach2, T. Netsch1, T. Schaeffter1, P. Börnert1, H. Eggers1, O. Lange3; 1Philips Research, Sector Technical Systems, Hamburg, GERMANY, 2Dep. of Electr. Eng. & Infor. Techn., Dresden Univ. of Technology, Dresden, GERMANY, 3TI 2 Computer Science, Technical University Hamburg-Harburg, Hamburg-Harburg, GERMANY. Introduction: High-resolution MR angiography requires accurate motion compensation to avoid image degradation by cardiac and respiratory motion. In contrast to conventional approaches, Hardy et al. [1] proposed a completely image-based technique, which uses neither triggering nor gating. In this work the completely image based approach has been extended to high resolution CMRA using sub-sampled high-resolution free-run radial images acquired using a steady state free precession (SSFP) sequence. Affine instead of a rigid body motion compensation was applied to register images selected automatically from a series of sub-sampled images. These transformed selected and subsampled images were combined to high-resolution CMRA images, in which the sub-sampling artifacts are mostly suppressed, and the SNR is improved due to averaging. Methods: The data are acquired using a radial SSFP sequence during a free-run acquisition (TR=3.2 ms, α=50°, SPIR-fat sat., 144 profiles, 192 samples, FOV=250mm) from healthy subjects on a 1.5T Philips scanner. The radial imaging is tolerant for sub-sampling artifacts. The fully adaptive weighted correlated averaging (FAWCA) approach [2] consists of an automatic pre-selection of 2D sub-sampled images, sub-pixel accurate registration, the combination of the selected interleaves to a high-resolution image and an adaptive weighted averaging (AWA) technique [3]. The sub-sampled images are generated by a sliding window reconstruction and image selection is based on cross-correlation techniques. An affine registration [4] instead of a multi-region rigid body registration is

used to correct the selected interleaves of the sub-sampled images. The AWA technique suppresses remaining uncompensated motion artifacts. Results and Discussion: In this work, a radial SSFP sequence has been utilized opposed to Cartesian acquisitions [1] or variable density spiral acquisitions [5], and high-resolution images of the RCA were successfully generated using the FAWCA approach. Furthermore affine instead of rigid body motion compensation techniques have been used which allow a more accurate image registration and a more efficient use of data. Respiratory motion as well as superimposed cardiac motion has been compensated as shown in Fig. 1. However, the streaking artifacts resulting from the significant radial k-space sub-sampling do not fully cancel out due to the affine registration. Conclusion: This work shows the basic feasibility of using subsampled high-resolution radial SSFP images to generate high-resolution CMRA images using the FAWCA algorithm. The use of affine registration methods can make more efficient use of data and still allow a real-time implementation. The use of motion compensation in combination with noise-robust filtering reduces residual motion artifacts significantly.

277 A quantitative comparison of algorithms for physiological artifacts correction M. Kilzer1, C. Windischberger1, E. Moser1,2; 1Institute of Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Department of Radiodiagnostics, University of Vienna, Vienna, AUSTRIA. Introduction: Physiological effects such as respiration and heart beat cause fluctuations in fMRI data which can be reduced using a number of post processing methods. Here we compare the physiofix algorithm[1] to the method of linear regression. Subjects and Methods: Respiration and heart beat are recorded synchronously during the acquisition of imaging data. The physiofix algorithm reorders the image data according to their relative temporal position in the physiological cycle. A Fourier series is fitted to the intensity time course of each voxel and subtracted to remove respiration/cardiac action related components. The method of linear regression fits the monitored physiological parameters to each voxel intensity time course using linear regression. The fit curve is then substracted from the voxel intensity. Both methods were applied in image space and k-space, in the latter case seperately to real and imaginary values as well as to magnitude and phase values. The test object consisted of four tubes with pump-generated pulsatile flow (1Hz) located around a static vessel. The physiological signal was assumed to follow the intensity in a voxel selected within one of the tubes. Measurement parameters were 128² matrix size, 1 slice, 4mm slice thickness, TE/TR=40/125, 960 repetitions. In addition, in vivo measurements were performed using a matrix size of 128², 3 slices, 4mm slice thickness, TE/TR=46/500, 270 repetitions. Results: The table below gives the maximum reduction of standard deviation percent. Furthermore, the figure shows the results for the test object correction in image space plotting the standard deviation

Flow Quantification: Motion and Artefacts after correction over the standard deviation of the original dataset, for physiofix (a) and linear regression (b). Discussion: We have shown that physiological artifacts can be reduced substantially using physiofix and linear regression. In general, physiofix results in superior artifact reduction, as it is more robust to non physiologically related signal fluctuation. However, it is a more complex algorithm and requires more user interaction to assign the physiological periods properly. Correcting real/imaginary values compared to magnitude/phase values gives similar results and does not require phase unwrapping. Correcting in k-space may be critical as sufficient SNR is required, and thus is limited to central k-space components. As such, spatial correlations might be introduced in the final data set [2]. Based on our results, applying the physiofix algorithm in image space seems to be the optimal choice. References: 1. Hu,MRM(1995) 2. Glover,MRM(2000) Acknowledgements: This work has been supported by the Hochschuljubiläumsstiftung of the City of Vienna (1472/2002) Max. Reduction PF k of Standard Deviation (%) [re/im]

PF k PF i [abs/phi]

LR k [re/im]

LR k LR i [abs/phi]

test object

66.4

48.5

66.1

67.7

45.5

71.1

heart beat in vivo

45.7

40.9

40.5

36.1

30.6

21.6

respiration in vivo

16.5

15.3

20.6

18.3

18.3

17.5

S153

278 Retrospective respiratory motion compensation for cardiac MRI A. Sigfridsson1,2, J. P. E. Kvitting1, L. Wigström1, M. Andersson2, H. Knutsson2; 1Clinical Physiology, Medicine and Care, Linköping University, Linköping, SWEDEN, 2Biomedical Engineering, Linköping University, Linköping, SWEDEN. Introduction: Cardiac MRI is known to be degraded by respiratory motion. Short scans can be performed using breath-hold techniques, while coronary artery imaging commonly use navigator gated sequences, acquiring data in a known static respiration position. Phase reordering reduces respiration artifacts without prolonging the scan, but may still leave some artifacts. Automatic motion compensation based on an image quality metric has previously been suggested [1,2]. In our approach we use a motion model and the respiratory bellow signal recorded during the scan to optimize the image quality. The parameters of a quadratic transfer function (from the bellow signal to the actual displacement) are optimized until a maximum image quality is obtained. Methods: In-vivo data were acquired on a 1.5 T scanner (GE Medical Systems, Milwaukee, WI). A sagittal 2D slice with frequency encoding in the SI direction was acquired using a gradient echo pulse sequence with the following parameters: matrix=256x256, FOV=32cm, slice thickness=7mm, flip angle=20°, TE=6.4ms, TR=20ms. Each k-space line was repeated 64 times, giving a total scan time of 5.28 minutes. Cardiac and respiratory information was collected simultaneously using a pulse oximeter and a respiration bellow. This information was then used for retrospective cardiac cine reconstruction to 64 time-frames using nearest neighbor interpolation. Pure translational motion in both in-plane directions was assumed, and a second-order polynomial transfer function from the respiratory data (obtained using the respiratory bellow) to translation in pixels was applied. A resulting image was reconstructed by adding a linear phase shift to the k-space lines, followed by regular IFFT. The parameters for the transfer function were obtained by gradient descent optimization using a local gradient entropy quality metric. Results: In-vivo data were obtained by scanning a 26 year old male volunteer, instructed to breathe normally during the experiment. A systolic time-frame was studied. The resulting images are shown below. The left image (a) shows the original image. The right image (b) shows the result after reconstructing the modified kspace. Improved delineation of the epi- and endocardium can be seen, and better demarcation of the peripheral pulmonary vessels.

Discussion: The method only compensates for in-plane motion in the 2D case, but extension to 3D will enable motion compensation in all directions. This might be useful for 3D flow acquisitions and high resolution morphological scans, such as coronary artery imaging. References: 1. Atkinson D. et al., [1997] IEEE Trans.Med.Imag. 16(6):903-910 2. McGee KP. et al., [1997] Radiology 205:541-545

S154

Short Communication Session

279 Comparative study of different state-of-the-art MRI sequences on the detectability and visibility of small tissues surrounded by air: a phantom study on 1.5T and 3T systems K. Smans1, H. T. C. Bosmans1, D. Bielen1, F. Maes2, P. Hamaekers1, R. Peeters1, G. Marchal1; 1Radiology, UZ Gasthuisberg, Leuven, BELGIUM, 2Radiology & Medical Image Computing (ESAT/PSI), UZ Gasthuisberg, Leuven, BELGIUM. Purpose: In order to find the most suitable sequence for MR colonography acquisitions after distension of the colon with air, we have performed a phantom experiment on both 1.5T and 3T MR systems. The purpose of the project was to compare the detectability and visibility of small tissues surrounded by air for state-of-theart MR sequences. Material and Methods: The sequences had to fulfill the requirements for MR colonography. This includes: acquisitions possible in a breath hold and high resolution to detect small lesions. As a model for polyps in the colon, we made a phantom that consisted of a layer of gel in which 8 groups of berries with different size and shape were plugged in. All berries were surrounded by air. The following sequences were compared on a 1.5T and 3T Intera systems (Philips, Best, The Netherlands) and on a 1.5T Sonata system (Siemens, Erlangen, Germany): 2D turbo spin echo (TSE), 3D ultrafast MR angiography (MRA), 3D SE, 3D TSE and a 3D truefisp (balanced TFE). A 2D SE sequence (band width 650 Hz/pixel) and a CT acquisition were considered as reference images. The amount of visible berries was counted and categorized as: none, some, most and all. The visibility of the berries was then judged as: totally invisible, faintly visible, well visible and excellent visibility. Results: This qualitative analysis was possible on all slices but a more quantitative approach would have been preferable and is under study. The following sequences failed to visualize the berries: 2D SE with low band width, TSE with echo train length (ETL) > 5 on the 3T system, 3D TSE and the balanced TFE on the 1.5T Intera system. The 3D MRA sequence provided excellent results on both the 1.5T and the 3T systems. On 1.5T, the TSE sequence, even with long ETL, performed better than the 3D SE sequence. The 3D SE sequence on 3T (with longer TR to reduce the specific absorption rate) was execellent and better than the same sequence on 1.5T. Truefisp was excellent on the Sonata system. Discussion and Conclusion: The 3T system has potentially more susceptibility artifacts but with appropriate sequences, the results are satisfactorily: the increased SNR seems to balance for susceptibility effects. Three-dimensional flash based techniques may be good candidates for screening after distension of the colon with air, even on the 3D system. On the 1.5T system, 2D TSE can also be recommended.

Short Communication Session 4:30 pm - 6:00 pm

Plate

280 Fiber tracking in the neonatal brain at 1.0T C. van Pul1, A. Vilanova1, G. Berenschot1, J. Buijs2, G. Roos2, P. Wijn1; 1Eindhoven University of Technology, Eindhoven, NETHERLANDS, 2Maxima Medical Center, Veldhoven, NETHERLANDS.

White matter structure can be visualized using fiber-tracking1. In neonates, white matter is less myelinated, resulting in lower anisotropy, which complicates fiber-tracking. We developed a fiber-tracking program and used it in 13 neonates. Diffusion tensor images were obtained using Pulsed Field Gradients in 6 directions with single-shot EPI on a Philips Gyroscan 1.0T (voxel=1.56x1.56x3mm). The tensor is calculated per voxel and interpolated. The main eigenvector is traced (line propagation). Stopping criteria are low anisotropy and maximum angle. A fiber-tracking image of a neonate is shown in Fig.1, arrows (A and B) indicating hypoxic-ischemic lesions. Note that in the hemisphere with lesion B, less fibers can be tracked, possibly due to irreversible structural damage. First results of fiber-tracking in the neonatal brain show that it is feasible, despite problems with low anisotropy, although only the larger fiber tracts can be followed. 1. Mori S et al. [2002] NMR Biomed. 15:1-14, 2002

281 Brain MRI signal changes in phenylketonuria A. Magenta Biasina1, G. Pompili1, C. Bonifacio2, R. Ciosci2, A. Zenga3, L. Fiori3, G. Cornalba2; 1Radiology department san Paolo hospital, Milan, ITALY, 2Institute of Radiology University of Milan, Milan, ITALY, 3Pediatric department san Paolo Hospital, Milan, ITALY. Purpose: To relate MRI abnormalities of CNS to clinical and biochemical findings in patients affected by classical phenylketonuria. Methods: From July 2002 to April 2003 an MR imaging study was performed in 20 patients (12-27 years old) with adequate and poorly controlled phenylalanine intake in order to define the abnormal findings and to relate these to clinical and biochemical findings. Results: MRI abnormalities consist in demyelination, with increased signal in FLAIR sequences, more evident in the periventricular deep cerebral white matter especially posteriorly in the optic radiations.

Short Communication Session Conclusions: MR imaging study confirmed the presence of abnormal signal of the white matter in all patients. The abnormalities were less severe in patients with adequate than in those with poorly controlled dietary treatment.

282 MRI invisible focal lesion detected by MRS, pseudonormalization of neonatal stroke E. R. Danielsen, K. Lundstrøm, C. Thomsen; University Hospital Rigshospitalet, Copenhagen, DENMARK.

S155

(LTSL) composed of the new 1.2-dipalmitoyl-sn-glycero-3-phosphoglyceroglycerol (DPPGOG) showed markedly enhanced circulation-time in-vivo (> 10h) and favourable Gd-release at 40-42°C. LTSL are shown to be a temperature-sensitive contrast-agent for MRI when gadolinium-chelate is trapped. At the therapeutic temperature of hyperthermia nearly 50% of the gadolinium is set free, and can be detected by T1-changes. Our results indicate that our system has the potentiality to improve monitoring of hyperthermia and thus the therapeutic outcome. LTSL may also be used for specific release of chemotherapeutics by moderate hyperthermia.

Pseudonormalization of DWMRI may occur early in neonates, Mader et al. (Stroke 2002;33:1142-45) reported a case where T2WMRI identified the DWMRI-invisible stroke. The present case had stroke negative MRI including T2WMRI. A previous healthy 4 weeks old boy presented with unconsciousness and seizures following gastroenteritis and hypoglycemia. He regained consciousness after 5-7 days, but remained hypotonic for several weeks. Extensive diagnostic evaluation could not reveal any aetiology. Neurologic examination was normal after two and five months.

284 Enhanced fractional anisotropy in the putamen with ageing: what does it really mean? C. Lavini, G. J. den Heeten; Academic Medical Center, Department of Radiology, Amsterdam, NETHERLANDS. T1-, T2-, DW-MRI and short TE MRS were performed at 12 days. MRS from the left basal ganglia was normal, but the mid-occipital spectrum was severely abnormal: NAA and Cr were 1/3, and Cho and mI were ½ of normal, lactate and lipids/macromolecules were elevated - MRS: stroke. MRI: Focal pathology negative. Lacking MRI visible focal pathology, an isolated mid-occipital spectrum could have been falsely interpreted as representing global injury and poor outcome.

283 Thermosensitive liposomes for MR-guided hyperthermia H. M. Reinl1, L. Lindner2, M. Peller1, P. Schneider2, N. Teichert2, H. Eibel3, R. Issels2, M. Reiser1; 1Dept. of Clinical Radiology, University of Munich, Klinikum Grosshadern, Munich, GERMANY, 2Clinical Corporation Group Hyperthermia, University of Munich, Klinikum Grosshadern, Munich, GERMANY, 3Max Planck Institute for Biophysical Chemistry, Goettingen, GERMANY. Hyperthermia has proven to be an effective treatment-concept for locally advanced deep-seated tumors. MRI could provide non-invasive temperature-control of local hyperthermia. Reaching the therapeutic temperature is a crucial factor for the outcome of the hyperthermic tumor-treatment. It has been shown that paramagnetic liposomes may support MR-thermometry. We demonstrate the feasibility of MR-thermometry using new temperature-sensitive long-circulating liposomes with encapsulated gadolinium-chelate GdDTPA-BMA. Long-circulating temperature-sensitive liposomes

Fractional Anisotropy (FA) maps provide quantitative measures of the tissues diffusion properties, and their averaged values in determined ROIs can therefore be compared between patient groups. While comparing white and grey matter ROIs in old and young volunteers in a DTI study, we came across marked, unexpected higher FA values in the putamens of the older group. Putamen is a small, low FA grey matter structure, surrounded by areas of very high anisotropy (corpus callosum and capsules). After volumetric analysis of the Putamen it appears that this finding might result from a putamen volume reduction (age related), with consequent larger partial volume effect in the slice direction (thickness=5 mm), rather than from a real FA change. This finding points out that a high slice-direction resolution is critical when examining FA in medium-to-small brain structures, and that volume changes might hide real FA changes or produce artefactual results.

S156

Short Communication Session

285

287

Hydrogen turnover in primary cultures of cortical cerebral astrocytes as detected by 2H-13C NMR A. Sierra1, S. Garrido1, M. Benito1, T. B. Rodrigues1, P. Ballesteros2, S. Cerdan1; 1Instituto Investigaciones Biomédicas, Madrid, SPAIN, 2Universidad Nacional de Educación a Distancia, Madrid, SPAIN.

Investigation of biexponential water diffusion in human cervical spinal cord T. Banasik1, A. Jasinski1, M. Hartel2, M. Konopka2, P. Pieniazek2, W. Weglarz1; 1Institute of Nuclear Physics, Krakow, POLAND, 2Silesian Diagnostic Imaging Center Helimed, Katowice, POLAND.

Introduction: 13C NMR studies provided important advances in our understanding of cerebral neurochemistry. However, 13C NMR methods are relatively slow and involve important assumptions. We recently showed that 13C(2H) NMR spectroscopy monitors adequately hydrogen turnover in 13C metabolites, a process occurring faster than 13C turnover. Here we address the hydrogen turnover of the lactate C3 resonance in astrocytes. Methods: Cortical astrocytes were incubated in KRB buffer containing 50% 2H2O and 5mM (1-13C) glucose. Aliquots from the incubation were analyzed by 13C NMR (1H and 2H decoupling). Results: The Figure shows the pattern of deuteration in lactate C3 with the 13C-2H isotopomers observed. Discussion: The exchange of the H2 and H3 protons are catalysed by different enzymes (LDH, PK, GAPDH). (2H)13C NMR analysis provides individual information on glycolytyc flow at the steps of the pathway where exchange occurs. Acknowledgements: Supported by strategic group grant of the Community of Madrid to PB.

Purpose: To investigate presence of the non-exponential diffusion in human cervical spinal cord. Materials and Methods: Imaging was performed on a GE SIGNA LX Echo-Plus scanner at Helimed, Katowice equipped with Research Mode using a modified DW EPI seqSuence. All DTI scans were gated using a peripheral pulse trigger. Axial DT images were acquired with a 64 x 64 matrix, FOV = 7 cm, slice thickness = 5 mm, NEX = 8 and b factor up to 7000 s/mm2. Results: Values of the fast and slow diffusion components were determined for different regions of the white (WM) and gray (GM) matter in directions perpendicular to spinal cord axis. For GM in anterior horns fast and slow componets were: Df=2.29*10-3mm2/s, Af=0.6, Ds=0.11*10-3 mm2/s, As=0.4 For WM in fasciculus cuneatus Df=1.43*10-3mm2/s, Af=0.5, Ds=0.04*10-3mm2/s, As=0.5. Conclusion: We have demonstrated non-exponential signal amplitude decay in human CSC for high diffusion weighting.

286 Localised proton spectroscopy of brain in a murine ESB model P. Meric1, C. Vidal2, J. L. Correze1, D. Dormont3, J. C. Beloeil1, B. Gillet1; 1CNRS, Gif sur Yvette, FRANCE, 2Institut Pasteur, Paris, FRANCE, 3CEA, Fontenay-aux-Roses, FRANCE. Purpose: To investigate the efficiency of MRI and MRS to estimate preclinical brain damage in prions diseases like ESB. Methods: Mice infected with an ESB strain (6BP1) were studied at a preclinical state in a 7T Varian Inova. Images were obtained from a bird cage probe (int.Ø= 39 mm) by T2 SE and diffusion sequences. Localized proton spectra were obtained by PRESS sequence in two voxels (3x3x3 mm) in frontal and posterior brain areas since it is known that the brain lesions at the clinical state are primarily localised in posterior brain areas in this model. Results: Proton spectra in posterior areas exhibited significant differences in infected mice with mainly a decrease in N-Acetyl Aspartate while no significant changes were found in images. Discussion: This strongly suggested that serial spectroscopic cerebral investigations of experimental models of prions diseases at preclinical states have to be undertaken.

288 Cardiac sarcoidosis - a diagnostic dilemma M. J. Michalak1, J. Walecki2, E. Michalak3; 1Central Clinical Hospital, Clinic of Cardiology, Warsaw, POLAND, 2Central Clinical Hospital, Department of Radiology, Warsaw, POLAND, 3Institute of Cardiology, Warsaw, POLAND. Authors presented interesting and difficult case of cardiac sarcoidosis in 53 years old female with 6 years history of histologically proven lungs sarcoidosis, who suffered from heart failure (NYHA III). The analysis of two consecutive transthoracic echocardiography showed an image that mimic cardiac infarction with akinesis of apical and anterolateral wall with decreased left ventricle ejection fraction (%LV EF) to 20 %. Coronary angiography was normal. Cardiac magnetic resonance (MRI) T2-weighted images demonstrated patchy infiltration located in the subendocardium in the area of apical and anterolateral wall of the left ventricle. Sarcoid infiltration was presented as linear zone of increased signal intensity which was more pronunced on T2

Short Communication Session weighted images. In T1-weighted, after gadolinum - DPTA, we didn't observed enhancement of the lesion. MRI allowed us to suspect sarcoidosis of the heart, which was histologically confirmed by biopsy of the heart.

289 A rare right atrial mass: Hydatid cyst D. Oncel, G. Oncel; Sifa Medical Center, Izmir, TURKEY. We report a right atrial hydatid cyst which is a very rare cardiac mass and represent less than 0.5 % of hydatid cystic disease. In the echocardiography of a 67 year-old woman, a mass in the right atrium was detected. For the further evaluation, a cardiac MR study was performed. A 1.5 T scanner (Symphony,Siemens) was used. In the T2 and T1W images, an 8 cm right atrial mass was observed. It was heterogenous showing mixed hypointense and hyperintense signal characteristics. In post-contrast studies, no significant signal intensity increase was detected. In TruFISP cine images, the mass showed no concurrent movement with right atrium, but the atrial filling and contractility were attenuated. The mass was reported as a calcific stage 4 hydatid cyst. The post-operative evaluation confirmed the radiological diagnosis. In conclusion, MRI is an effective imaging modality in identification of cardiac masses.

S157

291 Oxygen-enhanced parallel MR imaging of the human lung O. Dietrich1, C. Losert1, K. Nikolaou1, M. Peller1, S. O. Schoenberg1, B. Kuehn2, M. Nittka2, M. F. Reiser1; 1Department of Clinical Radiology, Klinikum Großhadern, Ludwig-Maximilians-University, Munich, GERMANY, 2Siemens AG, Medical Solutions, Erlangen, GERMANY. Purpose: To demonstrate the feasibility of integrated parallel acquisition techniques (iPAT) for oxygen-enhanced lung imaging in combination with an optimized respiratory and cardiac triggering scheme[1]. Subjects and Methods: We studied a healthy volunteer using an Inversion Recovery HASTE sequence[1] implemented on a 1.5T whole-body scanner (MAGNETOM Sonata, Siemens, Germany) with a dedicated 12-element iPAT surface coil and a GRAPPA reconstruction. We acquired a series of 80 images (20:air, 40:oxygen, 20:air). Results: The iPAT acceleration (factor 2) reduced the readout time from 193ms/slice to 103ms/slice. Figure 1 shows maps with the relative signal enhancement in the lung. We did not observe any artifacts due to the iPAT reconstruction

290 Cardiac NMR imaging of conscious hamsters using single-shot RARE E. Parzy1, D. Von Euw1, Y. Fromes2, P. G. Carlier1; 1Institut de Myologie, Paris, FRANCE, 2INSERM U582, Paris, FRANCE. There is increasing awareness not only that anesthesia offsets part of the benefit of studying tissue function and/or metabolism noninvasively by NMR, but also that more efforts should be devoted to NMR protocols in conscious animals. Circumventing anesthetics is particularly challenging for cardiac imaging. In this preliminary study, conscious hamsters were slipped into a 4.6 cm diameter cylinder, with the neck and legs immobilized. Half-Fourier singleshot RARE imaging was preceded with outer volume suppression to improve spatial resolution in the phase encoding dimension and with a double, selective and non-selective, inversion to reinforce black-blood contrast. Ungated double oblique short-axis views, 99.3 ms per image, were acquired in conscious hamsters, with an in plane resolution of 470*310 µm². Motion artefact-free images provided excellent visualization of cardiac anatomy, thereby demonstrating the possibility of fast cardiac NMR imaging in conscious rodents. By sorting piles of acquired images, end-systolic and enddiastolic volumes were approached.

Conclusion: Due to the reduction of acquisition time by about 50% we expect to be able to increase the number of acquired slices per respiration from 4 to at least 6 without prolongation of the acquisition. References: 1. Losert C et al. [2002] MAGMA 15(Suppl.1):123-124.

292 Infarcted recurrent phyllodes tumor or hematoma: a diagnostic challenge I. O. J. Verslegers, M. Van Goethem, W. Tjalma, C. Colpaert, I. Biltjes, I. Eerens, A. M. A. De Schepper; Universitair Ziekenhuis Antwerpen, University of Antwerp, Antwerp (Edegem), BELGIUM.

S158

Short Communication Session

Purpose: To document the MR mammography (MRM) findings in an infarcted phyllodes tumor. Methods and Materials: A 40-year-old woman presented with a palpable lump under a scar in her left breast. Ten months ago she had been operated for a phyllodes tumor. Recurrent tumor was suspected on the basis of clinical examination, mammography and sonography. MRM was performed on a 1.5T system with a doublebreast surface coil: axial T2-W turbo spin echo pulse sequence and coronal 3D FLASH-sequences before and after intravenous injection of gadolinium. Results: The MRM findings were consistent with a postoperative hematoma with granulation tissue. Surgery was performed. Histological diagnosis was an infarcted phyllodes tumor. Conclusion: This case illustrates the MRM appearance of a completely infarcted phyllodes tumor, that may mimick postoperative hematoma with granulation tissue. To the best of our knowledge, MRM findings of an infarcted phyllodes tumor of the breast have not been reported yet.

The 1.7s time resolution allowed Hepatic Perfusion Index (HPI) to be calculated from enhancement curves. HPI was significantly different in tumour and parenchymal regions. Reproducibility of HPI measurements between two identical studies was good. Incorporating separately acquired, full k-space data, using rigid registration and filtered constrained reconstruction , allowed tumour structure to be visualized at improved spatial resolution.

293 Phyllodes tumor of the breast T. Kanavou, M. Vlychou, K. Vassiou, A. Poultsidou, E. Athanasiou, K. Kappas, I. Fezoulidis; University of Thessaly, Larissa, GREECE. Herein is presented a case of phyllodes tumor of the breast that has been investigated with mammography, breast ultrasound and magnetic resonance mammography. A 57- year-old woman presented with a palpable mass in her left breast, which was freely movable. Mammography revealed a lobular mass with peripheral calcifications. Breast ultrasound revealed a 4x2.5 cm hypoechoic solid mass with peripheral cystic components and acoustic shadows. Magnetic resonance mammography at the T2 weighted images showed a lobulated mass with high signal and peripheral areas with no signal. After the administration of Gd the mass enhanced heterogeneously and gradually. The lesion exhibited a type I time-signal intensity curve and was considered benign. Histological report after surgical excision revealed a fibroepithelial lesion with features of phyllodes tumor and undetermined malignant potential without lymph node infiltration. Imaging investigation in such a lesion is useful not only in terms of preoperative assessment but also as follow-up.

294 Assessing reproducibility of HPI measurements in rapidlyenhancing liver metastases using fast central k-space MR acquisition M. J. White, D. Koh, L. Moore, D. Collins, M. O. Leach; Institute of Cancer Research, Sutton, Surrey, UNITED KINGDOM. Liver tumours often show rapid uptake of contrast medium, differentiable from parenchyma for only a short period (arterial phase), necessitating fast 3D dynamic imaging. These data form part of a reproducibility study utilizing reduced k-space sampling to achieve the required imaging speed. A patient with metastatic liver disease was dynamically imaged following injection of Gd-DTPA on two separate occasions (48 hours apart). 16 consecutive liver volumes with a temporal resolution of 1.7 seconds were obtained. Rapidly enhancing regions around metastases were seen; in some tumours differentiation from surrounding parenchyma lasted only 5-10s.

295 MRI value in detection and characterization of hepatic changes in Budd-Chiari syndrome I. Lupescu, R. Capsa, C. Zaharia, S. Georgescu; Hospital Fundeni, Bucarest, ROMANIA. Purpose: To present MRI findings in Budd-Chiari syndrome. Materials and Methods: 6 patients with clinical and ultrasonographic suspicion of Budd-Chiari syndrome, were explored between january 2001 and mars 2003 by MRI using a standard protocol and a 3D FSPGR multiphase acquisition after gadolinium injection.1 case was explored before and after double contrast injection (Resovist and Magnevist). Results: In all cases we have found a dysmorphic hepatomegaly with caudate enlargement and presence of ascites. The liver periphery present a moderately T1 low signal and high signal T2 relative to the central liver. Postcontrast in early phase, there was an inhomogeneous mottled enhancement on the central part of the liver. Hepatic venous thrombosis was demonstrated in all cases. Conclusions: MRI is a sensitive method to detect and characterize hepatic changes in Budd-Chiari syndrome. MR-angiography offer the possibility to better evaluate the thrombosis or fibrosis of hepatic veins and inferior vena cava.

296 Fine-needle biopsies in a conventional MR-scanner A. Nielsen; Aalborg University Hospital, Aalborg, DENMARK. Introduction: In some instances neither Ultrasound nor CT is the best choice for guiding a biopsy. Our purpose have been to develop an easy, but reliable method for estimating both X, Z and Y-axes needle position for use in a conventional MR-scanner. Methods: We have developed a cheap and easily available method,

Short Communication Session

S159

using tubes based on the surface of the patient, guiding the biopsy. We have evaluated the best possible sequences for guiding biopsies. We have been looking into the time-consumption in the MRscanner, the accuracy of the specimens and the complications. Results: 26 fine-needle aspirations and four biopsies have been performed. Estimated time has declined to approximately 30 min. and estimated needle-time of 2½ min. Conclusion: Use of a conventional MR-scanner release new guiding devices, is cheap and easily acceptable. Quality of biopsies seems as good as Ultrasound or CT-guided biopsies, and with few further risks to the patient.

297

298

Dynamic contrast-enhanced MRI at 1.5 T with Gd-DTPA measures angiostatic effects in tumors after treatment with anginex Q. G. de Lussanet1, R. G. H. Beets-Tan1, W. H. Backes1,2, D. W. J. van der Schaft1, J. M. H. van Engelshoven1,2, K. H. Mayo3, A. W. Griffioen1; 1University Hospital Maastricht, Maastricht, NETHERLANDS, 2University of Technology, Eindhoven, NETHERLANDS, 3University of Minnesota, Minneapolis, MN.

MRI based non-invasive virtual endoscopy of the small bowel J. B. Frøkjær, E. Larsen, E. Steffensen, A. H. Nielsen, A. Drewes; Aalborg University Hospital, Aalborg, DENMARK.

Effects of anginex on tumor angiogenesis were evaluated by dynamic contrast-enhanced MR imaging (DC MRI) at 1.5 T with Gd-DTPA [1]. C57BL/6 mice carrying B16F10 melanoma tumors were treated with anginex, TNP-470 or saline. Tumor growth curves and microvessel density (MVD) values were recorded to establish effects of treatment. DC MRI data were used for calculating tumor microvessel permeability surface area products (KPS) [1], which were used for comparison of the different treatment regimens. Anginex and TNP-470 treatments resulted in smaller tumor volumes [A] (p<0.0001) and lower MVD values [B] (p<0.0001) compared to saline. In accordance, anginex resulted in significantly reduced KPS values (64%, p<0.01) compared to saline [C], and TNP-470 showed a similar trend. This study demonstrates inhibiting effects of anginex and TNP-470 on tumor angiogenesis, by DC MRI with the use of a clinical MR system and clinical MR contrast agent. 1. de Lussanet, Q.G. [2003] Radiology in-press

Introduction: MRI is a new promising technique that allows imaging of the small bowel. The modality is shown to be exellent in description of mural and extramural small bowel pathology. This study is a supplement to sectional imaging (black lumen tecnique) for better evaluation of luminal ond mucosal changes. Methods: A non-invasive distension method with an aqueous solution of ispaghula husks and meglumine gadoterate (bright lumen) was performed orally over four hours prior to dedicated high resolution T1-FFE sequences for fast volume scanning. Volume rendered postprocessing (Philips 3D-endo software) allowed 3D virtual endoscopy. Results: One pilot study was performed with minimal discomfort, good luminal distension and demarcation, and low rate of artefacts. More than 80% of the small bowel was assessed by virtual endoscopy with fairly good quality an resolution. Conclusion: Virtual endoscopy is a new and interesting addition to sectional MRI of small bowel pathology.

299 Congenital Mesoblastic Nephroma: US and MRI appearances F. Serrao, V. Arcuri, P. P. Arcuri; Radiology Pugliese-Ciaccio Hospital, Catanzaro, ITALY. Congenital Mesoblastic Nephroma (CMN) (Bolande's tumour) is a rare renal tumour in childhood (3-6% of renal tumour) (1, 2). There have been only three previous reports concerning MRI imaging of CMN in the English-speaking literature (3). We present a further case diagnosed by US, confirmed by MRI. The US showed a tumour in right kidney upper pole, surrounded by a thin ipoechoic rim in which Color Doppler study revealed the

S160

Short Communication Session

typical "vascular ring sign". MRI was a determining factor providing additional information even about the particular histological form: CMN mixed form (typical and atypical) fig. 1. References: 1. Bisceglia M, Carosi I, Vairo M, Zaffarano L, Bisceglia M, Creti G [2000] Pathol Res Pract. 196 (3) : 199-204. 2. Kelner M, Droullè P, Didier F, Hoeffel JC [2003] Pediatr Radiol. 33: 123-128. 3. Puvaneswary M, Roy GT [1999] Australas Radiol. 43(4): 532534.

The axial and coronal T-1 and T-2 weighted sequences (with fat supression) were followed by contrast-enhanced MR-Angioraphy. A 33 years old female had a tumor of the labial and perineum region. There was an enlargement of the left leg. The MRI examination showed a large multiseptale vascular lesion of the left labial region with the typical signs of the vascular tumor. Furthermore there was a thrombus in the perineum region causing the vaginal stenosis. However, no arteriovenous malformation was found. In 6 years old female the size of the right leg was enlarged and the abnormal vascular structures were found. The MRI showed arteriovenous fistulas with vena femoralis dextra and vena saphena magna in the popliteal region. Besides, lymphovascular anomaly was revealed.

302 A pulse sequence for MR venography K. I. Gjesdal1, T. Storas1, J. Geitung2, A. Drolsum1; 1Ullevaal University Hospital, Oslo, NORWAY, 2Haraldsplass University Hospital, Bergen, NORWAY.

300 Combined MRI and spectroscopy targets prostate cancer in the central gland J. C. Vilanova, J. Barceló, M. Villalón; Clinica Girona, Girona, SPAIN. We present two patients with elevated PSA and previous biopsies negative for prostate cancer. MRI showed a low signal intensity lesion within the normal low signal intensity central gland from both patients, but 3-D spectroscopy (MRSI) could demonstrate a clear metabolic abnormality within these lesions. The subsequents guided biopsies confirmed prostate cancer. MRI/MRSI provides a noninvasive method to improve prostate cancer detection in selected patients not only in the well established peripheral gland but also in the central gland. Precisely, the central gland is an uncommon location for prostate cancer but also the most complicated region to be evaluated by purely morphologic MRI; as normal and pathologically central gland might appear with the same MRI signal intensity. It is important to evaluate the entire prostate gland (central and peripheral) whenever performing MRI/MRSI for prostate cancer assessment.

Throughout the last 10-15 years several centres have performed surgery, with venoplastics, of veins with regards to restore normal venous function. By introducing MRI we hope to both improve the diagnostics and to introduce one superior method for the entire diagnosis. We present preliminary by presenting one patient. Methods: Imaging took place on a 1.5T. The pulse sequence used for imaging the veins consists of a 3D multi chunk, multishot balanced transient field echo (B-TFE) with interleaved SPIR pulses. Key imaging parameters were TR=4.7ms, TE=2.35ms, FA=90°, slice thickness ~1mm, FOV=350-400mm. Results: The imaging method gave an excellent depiction of all insufficient veins and perforants, giving easy to use mapping for surgery. Discussion and Conclusion: This is a preliminiary result, but it is promising. With such an excellent visualisation of the veins and the possibilities of quantitative flow measurements, this may both improve and replace earlier diagnostic methods of venous insufficiency.

301 MR-angiography in diagnosis of Klippel-Trenaunay and Parks-Weber syndromes N. Gorbounov1, R. Forstner2; 1Novosibirsk State Medical Academy, Novosibirsk, RUSSIAN FEDERATION, 2Salzburg General Hospital, Salzburg, AUSTRIA. We studied two clinical cases: a 33 years old female with ParkesWeber syndrome and a 6 years old female with Klippel-Trenaunay syndrome. Patients were examined on 1.0-T system.

303 Intra articular form of osteoid osteoma- case report B. Hadzihasanovic, I. Ceric, M. Softic, F. Dalagija, I. Gavrankapetanovic, B. Gjikolli; University Clinical Center, Sarajevo, BOSNIA AND HERZEGOVINA. Purpose: Review of the rare form of intraarticular form of osteoid osteoma.

Short Communication Session Material and Method: MRI of the elbow were performed in the twenty one year old patient because of pain lasting more than one year. Discussion: Magnetic resonance revealed the typical finding. In the distal portion of right humerus an area of hypo signal in T1 was observed. After the intravenous administration of feromagnetic contrast media in T1 increased signal intensity. Diameter of the abnormality is 10 mm. Through the defect in cortical bone the abnormality is in close contact with thick sinovia. Conclusion: Magnetic resonance was seen superior in diagnostics of this rare form which has completely different radiology signs compared with osteoid osteoma of typical localization. Patient have been operated and pain stopped at the same day, typical sign for this benign tumor when the nidus has been completely extirpated.

S161

Acquisition with Relaxation Enhancement)1 sequence for use on a Varian Unity Inova 4.7T system. The SPARE sequence uses a train of spin-echoes and RARE phase-encoding along all three spatial axes. It samples a single point at the peak of each spin echo, eliminating the effects of phase shifts due to field inhomogeneities and chemical shift offsets. The resulting image exhibits no geometric distortions. The sequence is demonstrated using a wax phantom (Figure 1.) constructed with varying contrast areas and an embedded tissue section. The different regions are clearly distinguishable.

304 Different patterns of post-ischemic hyperemia identified by arterial spin labeling in mouse skeletal muscle D. Bertoldi1, D. Von Euw1, Y. Fromes2, C. Wary1, A. Leroy-Willig1, P. G. Carlier1; 1Institut de Myologie, Paris, FRANCE, 2INSERM U582, Paris, FRANCE. Arterial spin labeling (ASL) combined to NMR imaging allows multiple determinations of tissue perfusion. Dynamic investigations can be performed with high temporal resolution. This is of particular interest in skeletal muscles, post-exercise and /or post-ischemia. Using our own ASL methodology (SATIR), we investigated the influence of ischemic periods of increasing duration (3', 15' and 30') on post-ischemic hyperemia patterns (registered during 30'), in mouse calf muscles. Whilst short-term ischemia induced one single fast decaying hyperemia peak, longer ischemic periods resulted in longer and biphasic hyperemic responses (see figure). Between 10' and 20' hyperemia, the perfusion.time integral differed significantly (p<0.05) after 30' (507 ml.100g-1) compared to 15' ischemia (264 ml.100g-1) and 3' ischemia (85 ml.100g-1). To our knowledge, such quantitative perfusion data is not accessible to other techniques. It illustrates the role that ASL-based NMR techniques can play to investigate muscle vasodilatation control mechanisms in vivo.

305

References: 1. DJO McIntyre, F Hennel, PG. Morris [1998] j.man.reson. 130:58-62

306 T1 weighting by a set of regularly pulsed inversion pulses before imaging M. Braendli, G. Bongartz; Institute of Diagnostic Radiology, Basel, SWITZERLAND. A body area consists of various tissues; they each have their different T1 time. Generally spin or gradient echo sequences modify T1 contrast by adjusting TR and Flip angle. Furthermore T1 contrast may be altered by a single inversion recovery prepulse. We propose a simple strategy to prearrange longitudinal magnetization before imaging in such a way that longitudinal magnetization is inversely related to tissue's individual T1 time. Because signal is proportional to the longitudinal magnetization this strategy yields T1 weighting independent of TR and Flip angle. The method used is very simple: About six to seven regularly pulsed inversion pulses before imaging carry the system in a steady state in which tissue's individual longitudinal magnetization is inversely related to tissue's individual T1 time as postulated above.

307

3D SPARE imaging of wax embedded histological tissue sections M. O'Brien, D. J. O. McIntyre, J. R. Griffiths; CR-UK Biomedical Magnetic Resonance Research Group, St George's Hospital Medical School, London, UNITED KINGDOM.

In vivo NMR evaluation of electrotransfer induced damage to muscle A. Leroy-Willig1, M. F. Bureau2, J. Seguin2, D. Scherman2, P. G. Carlier1; 1Institut de Myologie, Paris, FRANCE, 2Pharmacologie Chimique et Génétique, FRE CNRS 2463, Paris, FRANCE.

We are investigating a solid-state MRI sequence for imaging the large archive of wax-embedded histological samples. Due to strong dipolar interactions solid materials exhibit short T2 relaxation times and strong line broadening, preventing the use of standard imaging sequences. We have developed a 3D SPARE (Single Point

Gene electrotransfer (ET) hold promise for muscle therapy and delivery of systemic proteins. Efficiency was evaluated by in vivo NMR with G-DTPA(1). For optimization of protocols, evaluation of tissue damage is also needed. We used a necrosis avid contrast

S162

Short Communication Session

agent, MnTPPS4, to detect ET-induced muscle necrosis. The ET protocol, known to induce tissue damage negligible at 200V/cm and significant at 300 V/cm, was applied upon rat tibialis and gastrocnemius muscles. MnTPPS4 0.07mM/kg was injected IV at 25 45 minutes after ET. NMR examination including T1 and T2 measurements was performed on a 4 T system, at 22-30 hours after injection. In treated muscles a heterogeneous T1-w signal increase was detected, by comparison with untreated muscles. T2 increased more than in muscle with ET at 200V/cm. T1 values ranged from 900 to1300 ms in the same locations, reflecting Mn uptake. References: 1. M. Paturneau-Jouas et al.Radiology in press.

308 Magnetic resonance techniques to study damage evolution in muscle A. Stekelenburg1, C. Oomens1, C. Bouten1, D. Bader2, K. Nicolay1; 1Eindhoven University of Technology, Eindhoven, NETHERLANDS, 2Queen Mary, University of London, London, UNITED KINGDOM. Pressure sores are localised areas of tissue breakdown in soft tissues. They are primarily caused by prolonged mechanical loading. It is not known how these external mechanical loads result in tissue damage. To investigate this we developed a rat model in which the tibialis anterior muscle is compressed between an indentor and the tibia. It was demonstrated in a previous study that T2weighted MRI is a promising alternative for histological examination to evaluate muscle tissue. We developed a MR-compatible loading apparatus in which we can apply pressure to the muscle and simultaneously collect MR data. The applied force will be measured using strain gauges and the deformation of the muscle will be determined with MR tagging. The development of tissue damage over time and the influence of reperfusion damage will be examined.

Figure 1 Experimental set-up (A). The indentor unit (B) can be placed in the set-up.

MR Angiography and MR Contrast Agents III / Small Animal Imaging and Spectroscopy III Sunday, September 21, 2003

Mini-Categorical Course MR Angiography and MR Contrast Agents III 8:00 am - 9:00 am

Willem Burger

309 MR Venography: techniques and indications E. Larsson; Department of Radiology, University Hospital, Lund, SWEDEN. MR venography (MRV) can be performed with several different techniques either without or with contrast injection. The 2D time-of-flight (TOF) method is performed perpendicular to the long axis of the veins, that is usually in the transverse plane, and is sensitive to slow flow. However, it is time-consuming, and tortuous vessels and collaterals with other flow directions are vulnerable to in-plane saturation effects. 2D or 3D phase contrast (PC) MRV can be performed using velocity encoding for slow flow. It can be performed along the vessels, but the 2D technique has limited resolution and the 3D technique is rather time-consuming. An advantage of the PC technique is the total background suppression. Steady state-free-precession (Balanced FFE, True-FISP) MR-imaging with high signal intensity of the blood-pool has recently emerged as a potential method for evaluation of the large veins without contrast injection. Contrast-enhanced 3D MRV is flow-independent and can be performed along the vessels. It relies on paramagnetic contrast agent induced T1 shortening for generation of intravascular high signal intensity. It can be performed with an indirect approach with contrast injection via the antecubital vein or a direct approach with diluted contrast agent continuously injected upstream into the venous territory under investigation. For the indirect approach with extracellular gadolinium based contrast agents, timing of the 3D gradient-echo imaging to the venous phase is necessary. If intravascular contrast agents (blood-pool agents) are used, timing is not needed and the scan times can be prolonged allowing higher signal-tonoise ratio and improved spatial resolution. The use of intravascular contrast agents also permits scanning of several vascular regions without a moving table technique. A drawback is the need for postprocessing because the contrast agent is present in arteries as well as in veins. The indications for MRV include detection of thrombosis or tumor involvement of the intracranial venous system, the large thoracic and abdominal veins and the veins of the extremities. The technique is also useful for the evaluation of patency of central veins in patients in whom difficulties are encountered to insert central venous catheters. MRV may also be used for the assessment of dysfunctioning hemodialysis arteriovenous fistulae.

310 New developments in contrast agents M. Laniado; Klinik und Poliklinik für Radiologische Diagnostik, Universitätsklinikum Carl Gustav Carus, Dresden, GERMANY. Purpose/Introduction: Magnetic Resonance Angiography (MRA) with contrast enhancement has become a clinically useful tool. It is

S163

the purpose of this paper to review the principles of the currently used contrast agents and to give an insight into the most recent developments of other contrast agents. Results: The extracellular gadolinium (Gd) compounds are safe enough to be bolus injected at dose ranges up to 0.3 mmol/kg of body weight and even higher. They are typically provided in 0.5 M concentration whereas a single agent (Gadovist, Schering, Germany) is be marketed in 1.0 M concentration. Of the extracellular Gd agents ProHance (Bracco, Italy) has the highest relaxivity due to weak protein binding in serum. Another approach to contrast enhanced MRA is the use of ultrasmall iron oxide particles (USPIO) of which Sinerem (Guerbet, France) is presently undergoing clinical investigation as a lymphographic contrast agent. The small superparamagnetic particles have considerable T1 relaxivity, long plasma half life (blood pool agent) and can therefore be applied for contrast enhanced MRA with T1-weighted gradient echo (GRE) sequences. The iron dose for sufficient contrast enhancement in plasma is relatively high (e.g. SHU 555 C, Schering, Germany, approximately 40 µmol/kg of body weight). Another type of blood pool agent is MS-325 (AngioMARK, EPIX Medical, U.S.A.). It is a Gd based agent that strongly but reversibly binds to albumin (80%-96% bound in human plasma) thereby exhibiting prolonged plasma elimination half life and increased relaxivity (dose range 0.01-0.05 mmol/kg of body weight). T1 relaxivity of MS-325 is approximately four to five times that of extracellularspace gadolinium-based compounds at 1.5 T and up to 10 times at 0.5 T. Increased relaxivity is due to the decreased tumbling rate of the albumin-bound chelate. MS-325 is eliminated primarily through the renal system. A second compound of the same class is B-22956/1 (Bracco, Italy). Finally, macromolecular contrast agents are under investigation for contrast enhanced MRA. One such compound is P792 (Guerbet, France), a macromolecular derivative of Gd-DOTA. It is a blood pool T1 agent that undergoes renal elimination. A second candidate is Gadomer-17 (Schering, Germany). The compound has carries 24 Gd chelates at the surface of a dendrimer. The excretion is via the kidneys. Discussion/Conclusion: The extracellular agents serve almost all purposes in contrast enhanced MRA. Of the blood pool agents, MS-325 shows the greatest clinical potential in terms of both safety and efficacy as well as its stage in clinical development.

Mini-Categorical Course Small Animal Imaging and Spectroscopy III 8:00 am - 9:00 am

Jurriaanse

311 Micro-MRI G. J. Strijkers; Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, NETHERLANDS. The availability of genetically modified small laboratory animals offers unique opportunities to measure the effects of an increase or a decrease in the expression levels of specific genes, or to investigate the effects of targeted gene mutations. The phenotyping of such models calls for reliable and non-invasive tools for the measurement of the consequences of the genetic modification. Traditional histological methods, however, are invasive and very time consuming. Furthermore they require special handling and

S164

MRI in Oncology: Diagnosis and Staging III

preparation of the specimen that cause loss of morphological structure. Micro-MRI is rapidly becoming an important alternative to traditional histology for rapid ex-vivo and in-vivo phenotyping, allowing the interrogation of intact, opaque organisms in three dimensions at relatively high resolution. In this mini-categorical course recent applications and developments of micro-MRI in biomedical research will be reviewed, mainly focusing on small laboratory animals. First, the physical limitations concerning spatial resolution, signalto-noise ratio, temporal resolution, et cetera, will be discussed. Advantages and problems encountered with non-invasive MR imaging of live animals will be addressed. Next, the efforts of several groups towards a digital MRI atlas of the mouse anatomy (“visible mouse” projects) will be reviewed. These MRI atlases can be compared to already published histological atlases, and used as a reference for future in-vivo studies. These studies usually involve fixated samples. The MR methodology and how contrast is obtained and may be manipulated will be explained. Finally, as an example of an in-vivo application, micro-MRI of the mouse cardiovascular system will be discussed. The beating mouse heart represents a formidable challenge, as the resting heart rate in the mouse may well approach or exceed 600 beats-per-minute, while MRI is very sensitive to motion. Gating the MRI scanning to the cardiac cycle, using the electrical ECG signal can effectively eliminate this source of motional blurring of the images. The technique has been used for the measurement of a variety of features of the transgenic mouse myocardium, during pharmacological stress and in models of heart failure. Significant progress is being made in vascular MRI of the mouse, which is primarily aimed at the detection and phenotyping of atherosclerotic plaques.

312 High-definition MRS W. Dreher; FB 2 (Chemie), Universität Bremen, Bremen, GERMANY. The recent methodological developments in localised in vivo MR spectroscopy (MRS) and spectroscopic imaging (SI) as well as the availability of improved hardware allow to speak of 'high-definition' MRS or SI to emphasise the possibility to acquire and process a greatly increased amount of information. MRS/SI measurements with increased spectral, spatial and temporal resolution and improved SNR can be performed on high field systems (4.7-11.7T) equipped with fast and strong self-shielded gradients and efficient RF coils. However, besides the hardware also the experiment and the data processing and evaluation have to be optimised. While multinuclear MRS/SI studies are carried out on different animals and organs, the majority of measurements use the proton signal and are performed on the brain. In single voxel MRS, a large number of metabolites can be observed and quantified simultaneously (e.g., ~ 20 at 9.4 T in rat brain) if pulse sequences with very short echo time, optimised shimming and sophisticated data evaluation techniques are used. If signals cannot be separated in 1D spectra, localised 2D MRS (J-resolved or correlation spectroscopy) offers a remedy, in particular for lower B0 strength and homogeneity. For applications on disease models, SI is often much more appropriate because spectra are simultaneously acquired in a large array

of small voxels. The SNR per unit measurement time (SNRt) and the minimum total measurement time (Tmin) are crucial for the practical use of SI methods. Tmin of classical SI is sometimes prohibitively long, even with modifications such as reduced k-space sampling, multiecho SI or if multislice SI is used instead of 3D SI. Therefore, various fast SI methods have been proposed which are spectroscopic variants of fast MRI sequences, e.g., EPI, spiral imaging, U-FLARE, BURST, GRASE, RARE or SSFP based imaging. However, the higher speed in k-space sampling should cause only minor (if any) SNRt losses. Which SI method should be applied depends on the specific application (metabolites of interest), the available hardware (B0, gradients, RF coils) and the required spatial, temporal and spectral resolution. Therefore, the availability of a large number of MRS/SI methods is essential to realise high definition MRS/SI. At high B0, MRS/SI using other nuclei than 1H (e.g., 13C, 31P, 15N) are of increasing interest for metabolic studies on animals. While such studies are still often performed in single voxels, SI is of increasing importance also for measurements with these nuclei.

Mini-Categorical Course MRI in Oncology: Diagnosis and Staging III 8:00 am - 9:00 am

Ruys

313 Node staging using USPIO-Enhanced MRI S. C. A. Michel; Diagn. Radiology, University Hospital Zurich, Zurich, SWITZERLAND. Accurate T- and N-staging is important for preoperative decision making in tumor patients. Whereas MRI is a reliable method for staging of most primary tumors, it is, like US and CT, less suited to assessing lymph node status. The major malignancy marker is a short-axis diameter exceeding 1 cm or 1.5 cm, respectively, depending on the lymph node area of interest. However, false-negative results for normalsize malignant nodes and the impossibility of differentiating enlarged benign hyperplastic from metastatic nodes impair staging performance. The continuing absence of a reliable imaging method leaves surgical lymphadenectomy as the effective but invasive gold standard for lymph node staging in most cancers. The use of dextran-coated ultrasmall superparamagnetic iron oxide (USPIO) particles may enhance the performance of MRI in noninvasive lymph node staging. USPIO is a lymph node-specific contrast agent in clinical development as a contrast agent for intravenous MR lymphography among other indications. After intravenous injection it extravasates into the interstitial space before being transported to the lymph nodes where, in healthy nodes, it is taken up by macrophages, resulting in signal loss – in contrast to a persistent signal received from metastatic nodes. Several clinical studies in head and neck, mediastinal, breast and pelvic cancer have confirmed that USPIO-enhanced MRI improves the detection of lymph node metastases.

Molecular Imaging: quo vadis? 314 MRS in oncology C. Arús; Departament de Bioquímica i Biologia Molecular, Universitat Autònoma de Barcelona, Cerdanyola del Valles, SPAIN. MRS of different nuclei (mostly 1H, 31P, 19F) has been acquired from human tumours (mostly from brain, prostate and breast), although most applications have concentrated on 1H data. Still, there are few applications for which MRS is routinely used as a diagnostic or prognostic tool as compared to the widespread use of MRI. There may be several reasons for that. MRS usually produces spectra, biochemical or biophysical information about the tissue, not morphological information like MRI. This spectral information is not easily usable by radiologists. To transform spectroscopic information (spectral pattern) into clinically relevant information, different approaches have been used with varying success. The simplest one, visual inspection of the spectral pattern, is presently suggested for diagnosing MRI suspicious breast masses, being the detection of a “choline” peak used as malignancy marker. Differential diagnosis between ring enhancing lesions (glioblastoma multiforme vs. abscess) as suggested by the detection of aminoacids, acetate and succinate is another paradigmatic example. Further sophistication involves calculating ratios of peak heights or areas. The best example of this may be the detection of prostate cancer from increases in the (Choline + Creatine / citrate) ratio. The single voxel (SV) approach can be extended to a multivoxel (MV) one, which is specially useful for heterogeneous tumours or whenever the abnormality zone is not easily defined from prior MRI exploration. Then, the approach used may transform the spectroscopic information into an image, either metabolic (maps of varying metabolite levels or ratios of them) as for prostate cancer diagnosis and staging, or nosologic (map of tumour/tissue type) for brain tumours. This last sophistication level usually requires some processing of the raw spectroscopic data, like feature selection and supervised/unsupervised pattern classification. Several groups actively pursue this approach to diagnose and grade brain tumours. Finally, the information produced by these various approaches needs to be displayed to the clinician in a user-friendly way. That is the field of the decision support systems. Then, for the case of SV, the spectroscopic pattern can be transformed by feature extraction procedures into a point in a 2D display (a graphical user interface) whose position represents the spectral characteristics of the case by placing it close to others of its kind and as apart as possible from other tumour types. An industrial prototype of such a system to help diagnose human brain tumours has been developed through a joint multicentric project (see http://carbon.uab.es/INTERPRET/ for further details).

S165

Plenary Session Molecular Imaging: quo vadis? 9:20 am - 10:50 am

Willem Burger

315 Molecular imaging of cell death to assess clinical diagnosis C. P. M. Reutelingsperger; Cardiovascular Research Institute Maastricht, Maastricht University, MD Maastricht, NETHERLANDS. Programmed cell death (PCD) is a well-organized biochemical process leading to the elimination of cells. PCD contributes to the homeostasis of the organism. Aberrancies of PCD may lead to the development and progression of diseases such as heart failure, tumor growth and neurodegeneration. Hence, measurement of PCD in tissues would provide important information about the disease. Such measurements should be preferably based on non-invasive procedures. PCD is characterized by a set of structural changes of the plasma membrane amongst which the expression of phosphatidylserine (PtdSer) is the most widely studied. Annexin A5 is a protein that binds with high-affinity to cell-surface expressed PtdSer. Thus it bears the potential of being a molecular imaging agent to measure selectivily cell death. We firstly established and validated Annexin A5 as a probe to measure PCD in vitro and in vivo in animal models. Next we successfully developed Annexin A5 into a molecular imaging probe to measure PCD in patients non-invasively using SPECT. Patients were studied who suffered from acute myocardial infarction, heart failure, atherosclerosis, or cancer. The results of these studies show that molecular imaging of cell death contributes significantly to the diagnosis. In addition, assessment of PCD using Annexin A5 provides knowledge about the efficacy of the therapy to treat the disease. This is of particular importance to the treatment of cancer. The Annexin A5 imaging protocol promises the possibility to evaluate a chemotherapeutic treatment shortly after its start. Clinical studies are underway to validate the Annexin A5 imaging protocol and to determine the benefit to the cancer patient.

316 Towards molecular imaging for therapy guidance C. Moonen; Laboratory for Molecular and Functional Imaging, University of Bordeaux 2, Bordeaux, FRANCE. Introduction: Molecular Imaging is a fast developing field with a large potential in diagnostics and therapy. Its basic definition was “the exploitation of specific molecules as the source of image contrast” (Weissleder), further adapted for medical applications as “the mapping of (patho)physiological and biochemical processes at the cellular and molecular level for diagnostics and therapy”. Molecular Imaging is a field that builds on the full deciphering of the human genome by looking at local gene expression levels and its consequences. The long-term vision of Molecular Imaging is that it will shift the current emphasis of health care on late diagnosis and treatment towards prevention, early diagnostics and (minimally-invasive) molecular therapies. Molecular imaging includes all imaging modalities. Objectives of Molecular Imaging: Early detection and characterization of disease (molecular signature) Understanding of its (patho)physiology

S166

Brain, Ischemia

Selection and evaluation of specific treatment Selected examples of Molecular Imaging: 1. Mapping of inflammatory activity has been demonstrated by imaging macrophage activity using USPIO contrast agents in atherosclerosis, ischemia, organ transplantation. 2. Tumor activity has been visualized by targeting neo-angiogenesis either by quantifying perfusion or other physiological parameters (see Thursday plenary lectures) or by mapping altered gene expression levels. Once specific markers are found, radiolabeling, or attachment of drug nanoparticles, can be used to locally deliver therapy. 3. Gene therapy can be guided by imaging. The delivery of a gene can be targeted using real-time image guidance. Then, expression of a co-expressed marker gene can be quantified using PET, MRI or other imaging modalities using labeled ligands. In addition, spatial and temporal control of transgene expression can be achieved based on a combination of MRI-guided local hyperthermia and a heat-sensitive promoter. 4. Stem cells may be used for image-guided cell repair. Stem cells have been isolated, and labeled with MR contrast agents, then re-injected and tracked towards their target using serial MRI. Such methods are promising in understanding the working of stem cells and may allow image-guided modification of stem cell therapy. Conclusion: Recent advances have demonstrated the potential of Molecular Imaging. Further significant challenges remain: 1. Develop surrogate imaging markers for major biochemical pathways (specific contrast agent design) and define clinically relevant imaging biomarkers of disease 2. Image effects of drugs on major pathways to aid drug development 3. Develop Image guided technologies for molecular therapy (local drug delivery, local gene therapy) 4. Demonstrate efficacy of new therapies

317 Molecular imaging and MRI: limitations and applications M. Hoehn; Department for Experimental Neurology, Max-PlanckInstitute for Neurological Research, Cologne, GERMANY. Molecular imaging using optical methods or also positron emission tomography (PET) has made quite some progress over the last decade. Although magnetic resonance imaging has proven to be able to produce highly resolved images with microscopic details, the applications were mostly recorded on histological sections where measurement time is amply available or were focussed on very small objects like plant stems where dedicated coils in high field systems produce the required sensitivity. However, the real test for the applicability of MRI as a microscopy tool comes if success with in vivo applications can be demonstrated, in competition with optical and PET imaging techniques. This is because here the ultra-high spatial resolution must be produced within limited recording time, forbidding endless signal averaging because the anesthetized animals will tolerate only a reasonable experimental time period. Furthermore, the focus of attention is of macroscopic dimensions, requiring corresponding hardware for extreme sensitivity. Recent advances with highly sensitive labeling techniques for cells now allow to produce a pronounced contrast of those labeled cells against the host tissue background into which they were brought. These labeling strategies require extensive biochemical and molec-

ular biological expertise as the label must efficiently enter the cell and remain in the cytosol for chronic experimental periods. Furthermore, toxicity must be guaranteed to be negligible to assure cell tolerance. On the MRI technological side, maximum sensitivity must be assured by developing dedicated detector systems for the particular objects and required resolution/sensitivity. This is seen in the context of minimized measurement time at the required spatial resolution (e.g. to observe small clusters of cells). Finally, the methodological basis must be designed and adapted for the special application needs with the physiological and well as MRI-technological requirements and achievable contrast and sensitivity of the particular application. After discussing the methodological boundary limits and requirements, the application will be discussed with the example of labeling cells for implantation in various organs, aiming at monitoring cell dynamics in vivo and following regeneration strategies using tissue grafts in the experimental animal.

Scientific Session Brain, Ischemia 11:20 am - 1:00 pm

Willem Burger

318 Post mortem cerebral diffusion measurements using MRI and NMR in neuropathologically analyzed brain tissue S. Brockstedt1, D. Topgaard2, M. Sjöbeck3, J. Lätt4, E. Larsson1, O. Söderman2, E. Englund3; 1Department of Radiology, Lund University, Lund, SWEDEN, 2Department of Physical Chemestry 1, Lund University, Lund, SWEDEN, 3Department of Pathology, Lund University, Lund, SWEDEN, 4Department of Radiation Physics, Lund University, Lund, SWEDEN. Introduction: NMR diffusometry can be used to classify water according to its mobility in tissue material. In diffusion weighted MRI investigations (DWI/DTI), the accessible b-range is often quite limited whereas in spectroscopic NMR, b-values which are orders of magnitude larger can be used. In this pilot study we present data from diffusion measurements of the post mortem brains acquired with MRI and spectroscopic NMR where spectroscopic diffusion measurements used b-values up to 1014 s/m2. Material and Methods: A case with clinically suspected brain white matter pathology was selected for post mortem cerebral MRI. The brain was fixed in 6% formaldehyde solution six weeks prior to investigation. Conventional MRI (T1W/T2W) followed by a DTI protocol was performed using a Siemens 3T Allegra MR scanner. Spectroscopic NMR diffusometry was later performed on samples from frontal white matter and the posterior part of corpus callosum, using a Bruker DMX 200 spectrometer operating at a proton resonance frequency of 200.13 MHz. Water diffusion was followed with the pulsed field gradient stimulated echo method using gradient pulse lengths of 5 ms and eight effective diffusion times in a geometric sequence between 10 ms and 1 s. The gradients were incremented in 64 steps between 1 mT/m and 10 T/m, covering a q-range of 200 - 2×106 m-1. A neuropath logical analysis of the examined tissue was subsequently performed. Results: MRI showed evidence of various degrees of white matter alterations also identified on histopathological examination. The degree of white matter pathology correlated with gradually lower fractional anisotropy and increased ADC (Fig 1).

Brain, Ischemia Neuropathologically, the analysed white matter showed ischemic alterations, including a mild edema. Spectroscopic NMR data showed the presence of several classes of water with different diffusion behaviour (fig 2). In general, we detect “free” water with a diffusion coefficient close to bulk water and a second class which shows a slower diffusion. Finally, a minor component of nonGaussian diffusion and a root means square displacement less than 1µm independent of diffusion time was detected. This class of water is thus confined to a compartment size in the order of µm. Discussion: Diffusion MRI has become a valuable tool for investigations of different diseases affecting cerebral white matter. The additional information that can be obtained from spectroscopic NMR diffusion measurements may add quantitative information regarding water mobility in different compartments (e.g. extra- and intracellular water, i.e. edema of different types).

S167

mated segmentation software was developed. Results: After surgery one major and one minor complication was registered and after stent placement one patient suffered a minor stroke. In 3 of the 44 performed pre-treatment DWI-series 6 hyperintense lesions were appreciated. Seventeen new hyperintense lesions were recorded on the post-treatment DWI-series after surgery in 4 of the 28 procedures (14%) and 42 new hyperintense lesions after endovascular treatment in 13 of the 26 procedures (50%). In 13 cases all new lesions were located in the vascular territory of the carotid artery on the ipsilateral side of the intervention. Four endovascular treated patients also showed 10 new hyperintense lesions in the contralateral side including 2 patients with 6 new lesions in the posterior circulation. The number of lesions after intervention varied from 0-13 after surgery and 0-10 after stenting. The volume load on the DWI after intervention varied between 0,31 cm and 4,74 cm³ (median 0,43 cm³) in the endarterectomy group and 0,01 cm³ and 9,72 cm³ (median 0,09 cm³) in the stent group. Conclusion: Small ischemic lesions could be revealed by diffusion weighted MRI. Ischemic events were detected more often after carotid stenting than after endarterectomy. Most of the detected lesions did not cause neurological deficit and were clinically silent.

320 Mismatch between lactate and ADC changes in acute stroke A. Stengel1, T. Neumann-Haefelin1, H. Lanfermann2, O. Singer1, C. Neumann-Haefelin3, F. Zanella2, U. Pilatus2; 1Department of Neurology, JW Goethe University, Frankfurt am Main, GERMANY, 2Institute of Neuroradiology, JW Goethe University, Frankfurt am Main, GERMANY, 3Lead Optimization, Aventis Pharma GmbH, Frankfurt am Main, GERMANY.

319 Detection of ischemic lesions with diffusion-weighted MRI after stent placement or surgical endarterectomy for symptomatic atherosclerotic disease in the carotid artery H. Flach1, M. Ouhlous1, J. F. Veenland1, M. R. H. M. van Sambeek, PhD2, L. C. van Dijk, PhD1, A. van der Lugt, PhD1; 1Radiology, EMCR Dijkzigt, Rotterdam, NETHERLANDS, 2Vascular Surgery, EMCR Dijkzigt, Rotterdam, NETHERLANDS. Purpose: To determine the incidence of ischemic events after carotid stenting or endarterectomy with diffusion-weighted imaging Subjects and Methods: From February 2001 till November 2002 diffusion-weighted imaging (DWI) was performed in 54 patients with high grade symptomatic carotid stenosis within 48 hours after carotid endarterectomy (n=28) or carotid stent (n=26) placement. In 44 of these patients DWI was performed also before treatment. MR Imaging of the brain was performed with a 1,5T scanner (Signa, General Electric) and a dedicated head coil. The MRI protocol consisted of a DWI sequence with a slice thickness of 5 mm and b-value of 1000 s/mm2 (SE/EPI, TR 12999, minimum TE, FOV 24x19cm, matrix 128x160, 3 NEX). Combined diffusionweighted images were generated and analyzed. The number, location (ipsilateral versus contralateral) and volume (cm³) of new hyperintense lesions were assessed by two experienced radiologists. To calculate the volume of the lesions on DWI a semi-auto-

Introduction: The DWI-PWI-mismatch has been suggested to indicate salvageable tissue at risk of infarction in the acute stroke. However, the exact risk of this tissue is often difficult to determine. To provide additional information about the status of the hypoperfused tissue, we explored the potential utility of metabolic markers of anaerobic glycolysis (Lactate, Lac) and neuronal integrity (Nacetylaspartate, NAA) using turbo spectroscopic imaging (TSI). Methods: Nine acute stroke patients were examined within 24 hours of symptom onset. In a subset of patients follow-up examinations were done at day 1 or 7 to assess final infarct size. The MRI examinations were performed on a 1.5 T Philips Gyroscan Intera scanner using standard parameters for the conventional stroke sequences (including DWI, PWI, T2W). For the TSI experiments, four individual phase encoded spin-echoes were acquired with the following acquisition parameters: TR = 1500ms, TE = 144, 288, 432, 576ms, FOV 240mm, slice thickness 15mm, 32 x 32 phase encodings, 2250Hz bandwidth, 256 data points. The total measurement time was 6 minutes. Volume preselection (PRESS) and multiple-slice outer volume suppression (OVS) were used to suppress lipid signals from the skull. The spectroscopic method was validated in phantoms and healthy volunteers. Effects caused by J-coupling (Lac) were shown to be negligible. Peak integration was performed in the time domain with the csxsoftware [1]. To allow the comparison of signal intensities between subjects, coil loading corrections were applied. Results: In n=8/9 acute stroke patients we found regions with increased Lac and different degrees of NAA reduction, clearly extending beyond the borders of diffusion restriction (fig1). Within the DWI lesion, we consistently found increased lactate and

S168

Brain, Ischemia

reduced NAA, even in patients presenting only few hours after symptom onset. Discussion: We observed lactate abnormalities extending clearly beyond the regions with abnormal diffusion, probably indicating regions at risk of infarction within the DWI-PWI-mismatch. References: 1. P Barker. JHU Baltimore. MD. Fig.1 Signal intensities plotted against the ADC values in nine acute stroke patients (25 voxel/patient). Rectangle: healthy tissue.

Results: This study demonstrates, for the first time, a wide gradient of ischemia-related metabolic anomalies within the abnormal area delineated by DWI during AIS. In the narrow range of decreased mean ADC values (0.6 to 0.40 x10-9 m2.s-1), a 33% decrease in mean ADC is associated with a 122% increase in Lac / NAA ratio. Mean ADC values never fall under 0.40 x10-9 m2.s-1 within the severely affected ischemic tissue while SI still detects a large metabolic heterogeneity inside areas showing similar decreased mean ADC values close to this threshold. This metabolic heterogeneity inside the area of severely decreased mean ADC is clearly illustrated using SI, whether or not a diffusion-perfusion mismatch is present Discussion / Conclusion: Our results indicate that the region of very low mean ADC values observed during AIS contains areas of various tissue damage intensity characterized by SI in relation with different stages of cellular metabolic injury. These data demonstrate that the lowest ADC values observed during the hyperacute phase do not correspond to maximal tissue injury and, so far, highlight the limitations of DWI to evaluate the severity of cellular suffering. This observation may explain why ADC mapping does not reliably predict the final infarct size.

322

321 The metabolic counterpart of decreased apparent diffusion coefficient during hyperacute ischemic stroke. A brain proton MR spectroscopic imaging study F. Nicoli1,2, Y. Le Fur1, B. Denis1,2, J. Ranjeva1, S. Confort-Gouny1, P. J. Cozzone1; 1Centre de Résonance Magnétique Biologique et Médicale, UMR CNRS 6612 Faculté de Médecine, Marseille, FRANCE, 2Neuro-vascular Unit, Hôpital Sainte-Marguerite, Marseille, FRANCE. Purpose/Introduction: Recent studies have shown that the brain ischemic area defined by the map of decreased apparent diffusion coefficients (ADC) obtained by diffusion-weighted imaging (DWI) during the first hours of ischemic stroke includes a significant part of ischemic penumbra. We hypothesize that the misjudgement of final infarct size by ADC mapping may be related to a restricted ability of DWI to capture variations in the intensity of cellular suffering. In an attempt to characterize metabolically the hypoperfused brain parenchyma, we studied the relationship between ADC values and brain metabolic parameters measured by proton MR spectroscopic imaging (SI). Subjects and Methods: Six patients with hyperacute ischemic stroke (AIS) were explored within the first 7 hours following onset, using a MR protocol including T2*-weighted MRI, DWI, SI, perfusion-weighted imaging and MR angiography. Four out of the six patients were explored within 6 hours (h) of onset (3h30, 4h30, 5h and 6h respectively). They had proximal middle cerebral artery occlusion without cervical artery stenosis or occlusion. For the two other patients who were explored respectively within 4 and 7 hours of onset, MRA showed no intra or extra-cranial artery occlusion.

A study of the temporal evolution of water diffusion parameters in grey and white matter in human ischaemic stroke S. Muñoz Maniega1, M. E. Bastin1, A. J. Farrall2, P. Hand2, V. Cvoro2, C. S. Rivers2, P. A. Armitage2, M. Dennis2, J. M. Wardlaw2; 1Medical and Radiological Sciences (Medical Physics), University of Edinburgh, Edinburgh, UNITED KINGDOM, 2Clinical Neurosciences, University of Edinburgh, Edinburgh, UNITED KINGDOM. Purpose/Introduction: The purpose of this study was to investigate whether there was any difference in the temporal evolution of mean diffusivity () and fractional anisotropy (FA) values of grey and white matter after stroke. Subjects/Methods: Twenty-one patients with moderate to severe cortical ischaemic stroke underwent diffusion imaging (19 diffusion tensor and 2 diffusion-weighted) at 0-1, 5-7, 10-14, 30 and 90 days after stroke. None of these patients received neuroprotective or thrombolytic drugs. Imaging data from the last four time points were co-registered to the first time point. Outlines were traced on all slices around the hyperintense lesion on the first time-point DW images. Using the T2-weighted images, multiple small circular regions-of-interest (ROI) were placed in grey and white matter within the lesion and in comparable contralateral tissue. Values of {grey,white} and FA{grey,white} were measured in these ROI at the five time points. To improve comparisons between individuals, the ratio of and FA values in the lesion and normal contralateral brain, r and FAr, were calculated. To ensure that within patient effects, rather than inter-patient variability were tested, differences between r and FAr values of grey and white matter were assessed using a two-tailed Student's paired t-test with p < 0.05 being considered significant. Results: The values of r and FAr for grey matter showed different patterns of change to that in white matter. Both decreased rapidly 0-1 days post ictus, with r{grey} = 0.80, r{white} = 0.71, FAr{grey} = 0.95 and FAr{white} = 0.80. Between the first and second time points r{grey} remained constant, while r{white} started to increase (0.71 to 0.82). During the same period FAr{grey} decreased only slowly (0.95 to 0.91), while FAr{white} fell more rapidly (0.80 to 0.55). Thereafter,

Brain, Ischemia r{grey,white} values rose slowly over time, while FAr{grey,white} values remained depressed. At the first and third time points r{grey} was significantly greater than r{white} (p < 0.01). At all other time points r{grey} was approximately equal to r{white}. At all time points FAr{grey} was significantly greater than FAr{white} (p< 0.01 ). Discussion/Conclusions: This is the largest study to date investigating the temporal evolution of water diffusion parameters separately in grey and white matter after stroke. The results show that the evolution of grey and white matter and FA values differ. Future work is required to determine whether characterizing these differences are useful in predicting stroke progression, severity and outcome.

323 Characterization of ischemic damage in neonates using diffusion anisotropy indices C. van Pul1, M. Janssen1, J. Buijs2, G. Roos3, P. Wijn1; 1Applied Physics, Eindhoven University of Technology, Eindhoven, NETHERLANDS, 2Neonatology, Maxima Medical Center, Veldhoven, NETHERLANDS, 3Radiology, Maxima Medical Center, Veldhoven, NETHERLANDS. Introduction: ADC measurements are valuable as early indicator of ischemia in the neonatal brain1, but it remains unclear whether anisotropy indices give additional information about changes in tissue structure. We have evaluated three well-known anisotropy indices (FA2, RA2, Cl3) for characterizing ischemic and hemorrhagic-ischemic lesions in neonates. Subjects and Methods: DTI images were obtained using Pulsed Field Gradients with 3 b-values (0, 400 and 800 s/mm2) in 6 directions with single-shot EPI with voxel size of 1.2x1.2x4mm (Philips Gyroscan 1.0T). For post-processing, Mathematica® was used. The anisotropy indices were tested in 16 volunteers and used in 16 neonates (6 with ischemic white matter (WM) infarcts (I-WM), 4 with hemorrhagic-ischemic infarcts (HI-WM) and 6 with normal MRI). Diffusion information was evaluated in a Region Of Interest in frontal WM (volunteers) or in lesion (L) and contra-lateral (C) WM. Results: ADC values in neonatal WM are larger than in adult WM and all anisotropy indices are significantly lower (Table 1, data presented as mean ± standard deviation). FA is relatively large compared to other indices and shows a smaller relative error. Values for Cl in WM are low and hardly distinguishable from isotropic. Table 1

ADC Mean ±σ

FA Mean ±σ

RA Mean ±σ

Cl Mean ±σ

Adult WM (n=16) 881 ± 50

0.43 ± 0.05

0.26 ± 0.05

0.18 ± 0.05

Neonatal normal

1716 ± 247

0.27 ± 0.09

0.16 ± 0.05

0.10 ± 0.04

Neonatal I WM

L: 793 ± 242;

L: 0.33 ± 0.08; L: 0.20 ± 0.05; L: 0.12 ± 0.03;

(n=6). L=lesion;

C: 1409 ± 175 C: 0.21 ± 0.05 C: 0.13 ± 0.03 C: 0.08 ± 0.03

WM (n=6)

S169

Discussion and Conclusion: Although anisotropy in the neonatal brain is lower, due to less myelination of white matter1, relative changes in all three anisotropy indices can be used to characterize WM lesions in addition to ADC. All indices are increased, with no significant difference between ischemic and hemorrhagic-ischemic lesions, although standard deviations in HI-WM lesions are larger, which might be due to blood (clot formation and susceptibility effects)5. The relative error in Cl and RA is larger than in FA. We conclude that FA is more suitable than RA and Cl for characterizing lesions, although the additional information of anisotropy for prognosis of neurological outcome still has to be investigated. References: 1. Neil J et al. [2002] NMR Biomed. 15:543-552 2. Basser PJ et al. [1996] J. Magn. Reson. B. 111:209-219 3. Peled S et al. [1998] Brain Res. 780:27-33 4. Sotak CH [2002] NMR Biomed. 15:561-569 5. Atlas SW et al. [2000] AJNR 21:1190-1194

324 Volume exchange during cerebral ischemia: the interaction of the diffusion components W. J. Gartner1,2,3, T. P. Duprez1, C. B. Grandin1, F. Peeters1, G. Cosnard1, R. Kleinert3; 1Department of Medical Imaging, Cliniques Universitaires Saint-Luc, Brussels, BELGIUM, 2MRIProject, Technical University, Vienna, AUSTRIA, 3Division of Neuropathology, Karl-Franzens-University, Graz, AUSTRIA. Introduction: The splitting of the inner-volume of diagonal excitations on the basis of parameters used for multicomponent curve fitting has been described [1-3]. Invariant tetrahedral diffusion directions provide a measure of the orientation-dependent diffusion. Subjects and Methods: Eight stroke patients who were considered for thrombolytic therapy underwent admission and follow-up MRI. Imaging involved fast FLAIR, 3D time-of-flight, gradient-echo EPI, fat-suppression in single-shot diffusion spin-echo EPI, and a 6.25 kHz bandwidth of diffusion sensitized line scans at b-factors up to 4500 s/mm2 [4, 5]. Geometric mean values were calculated for selected columns of the 128x64x10x4 data field to get an average of the volume content. With the parameters F, ξ and η, two asymptotes

C=Contra-lateral Neonatal HI WM

L: 1270 ± 328; L: 0.29 ± 0.19; L: 0.18 ± 0.13; L: 0.11 ± 0.09;

(n=4). L=lesion;

C: 1626 ± 289 C: 0.16 ± 0.06 C: 0.09 ± 0.03 C: 0.06 ± 0.01

C=Contra-lateral

ADC is significantly decreased in ischemic lesions compared to contra-lateral WM and decreased (not significantly) in hemorrhagic-ischemic lesions. For all three anisotropy indices, the relative increase in lesions is about 50% (Fig.1), both in ischemic lesions (similar as in adults4) as in hemorrhagic-ischemic lesions, although in the latter, standard deviations are larger.

satisfied the diffusion decay. Variation of the slow ADC2 and the fast ADC1 values justified new constraints by keeping the fast component unchanged at a variable water exchange rate.

S170

Post-processing and Quantification: fMRI and MRS

Fig. Multi-infarct lesions at actual, 1st and 8th day after sudden onset of Wernicke aphasia. A Left PCA stenosis in TOF. B FLAIR showing gliotic scars. C MTT-mapped relative tissue perfusion. D-F Low b-factor LSDI. G-I High b-factor LSDI. J-L Fast-diffusing water fraction. M-O Slow-diffusing water fraction Results: Lowered fast and lowered slow ADC values attributed ischemia to hyperintense image areas. Bolus-tracking perfusion MRI identified hypoperfused territories. While organelles became pyknotic, the slow-diffusing volume expanded because of microvacuolation of the perineuronal processes (Figure). Depending on the lowering of the slow component the relative volume percentage increased despite recovery of the fast one. Hypersignal in FLAIR resulted from a net increase of the water amount. Leukoaraiosis became separated from cavities lined by astroglia with only fast-diffusing fluid. The lowest exchange rate was associated with encephalomalacia. Fast exchange of intracellular volume correlated with salvageable tissue. The standard deviations of the orientation-dependent diffusion coefficients focused on tissue necrosis. Elimination of signal from glycogen and lipid allowed diagnostic relevant volume ratios to be derived from the images. Conclusion: Multicomponent diffusion imaging exhibits a distal and a proximal lesion volume which are linked to each other. Water exchange is determined by respective volume size. Analysis of the exchange rate is capable of classifying the lesion volume. References: 1. Mulkern RV et al (1999) NMR Biomed 12:51-62 2. Gartner WJ (2002) ISMRM Workshop on Diffusion MRI Syllabus:200-203 3. Sehy JV et al (2002) Magn Reson Med 48(5):765-770 4. Grandin CB et al (2002) Radiology 223:361-370 5. Gartner WJ et al (2002) MAGMA 15(Suppl 1):144-145

325 Contrast Enhanced MRI combined with epiaortic and intracranial MRA for the assessment of latent brain barrier damage in candidates for Carotid Endarterectomy M. Caulo, A. Tartaro, D. Di Giorgio, M. Mansour, A. Di Iorio, C. Colosimo, L. Bonomo; Dip. Scienze Radiologiche, Università G. D'Annunzio, Chieti, ITALY. Purpose: Cerebral hyperperfusion syndrome (CHS) is a serious complication of carotid endarterectomy (CEA) caused by loss of cerebral autoregolation, resulting from chronic cerebral ischemia. The aim of this study was to evaluate latent hypoxic damage to the blood brain barrier (BBB), as a consequence of chronic cerebral

ischemia, by means of contrast-enhanced brain MRI combined with epiaortic and intracranial MRA. Materials and Methods: 589 patients (mean age 69.8 years), considered candidates for CEA, underwent in a single session brain MRI before and after the iv administration of 0.1 mmol/kg gadobenate dimeglumine (Gd-BOBTA); MultiHance) and MRA of epiaortic and intracranial arteries. None of the patients had a history of recent stroke. BBB breakdown was assessed as areas of contrast enhancement on T1 weighted MR images with or without signal alteration on FLAIR images. Results: Twenty-seven areas of contrast enhancement were found in 21 (3.5%) patients; all but two of these areas were hyperintense on FLAIR images. Combined severe stenoses of the internal carotid (ICA) and intracranial arteries were found in 15/21 patients (71.4%) whereas Isolated stenosis of the ICA or intracranial arteries was found in 4/21 (19.0%) and 2/21 (9.5%) patients, respectively. Conclusions: Combined contrast-enhanced brain MRI and MRA, performed in the workup before CEA, can reveal BBB breakdown in patients with chronic cerebral ischemia and arteriosclerosis of intracranial and epiaortic vessels.

Scientific Session Post-processing and Quantification: fMRI and MRS 11:20 am - 1:00 pm

Jurriaanse

326 Multifractal analysis of fMRI data Y. Shimizu1,2, C. Windischberger2, S. Thurner3, E. Moser2,4, M. Barth2,4; 1NuHAG, Department of Mathematics, University of Vienna, AUSTRIA, 2Agnmr, Department of Medical Physics, University of Vienna, AUSTRIA, 3Department of Ortorhinolaryngology, University of Vienna, Medical School, AUSTRIA, 4Department of Diagnostic Radiology, University of Vienna, Medical School, AUSTRIA. Introduction: We investigated the potential of a wavelet-based multifractal methodology to distinguish different patterns in timeseries of functional Magnetic Resonance Imaging (fMRI). Based on the results we introduce a novel fractal parameter for the evaluation and visualization of fMRI data, where no a priori information on the paradigm is required. Methods: The Continuous Wavelet Transform (CWT) is basically a convolution of the signal with shifted and dilated versions of the mother-wavelet. At small scales the CWT shows high time resolution. This allows for tracking signal intensity variations by observing the behaviour of the CWT-coefficient-maxima at different scales. We quantified the strength of the singularities in terms of the Hoelder exponent[1]. The higher this exponent the more pronounced the intensity change. We extracted the multifractal spectrum of the cumulative sum of each voxel time series, according to a procedure known as the Wavelet Modulus Maxima Method (WTMM)[1]. The multifractal spectrum reveals all Hoelder exponents h in the time series examined along with their frequency of appearance D(h). Geometrical characteristics of the latter, height Dmax, most frequent exponent hmax, as well as Full-Width-atHalf-Maximum (fwhm) were considered to allow for quantitative comparison of spectra. These features were combined to a single

Post-processing and Quantification: fMRI and MRS empirical parameter Pc: Pc = (hmax/Dmax)*fwhm. Using this analysis we evaluated a set of hybrid data, produced by superimposing a sequence of haemodynamic response functions on the time series of resting state fMRI data in an area of 5x5 voxels. Results: The figures presented show that the activated region can be distinguished from the non-activated by means of all three parameters (i.e.,hmax,Dmax,fwhm),but only the combination of all three values singles out the activation to a high degree. The differences of the median of activated and non-activated voxels was calculated to 0.2, 0.1, 0.2 and 0.5 for hmax,Dmax,fwhm and Pc, respectively. We additionally observe differences correlating with anatomical structures (i.e.,gray/white matter, see figures), which suggest this method for tissue segmentation.

Fig. 1: Parameter plots obtained by geometrical evaluation of multifractal spectra.

Fig. 2: Spectra for voxels in different brain areas. Discussion: Event-related signal changes as well as physiological effects in fMRI time series could be detected successfully and quantified by the fractal index Pc. This analysis method does not require any prior knowledge about the paradigm and provides a potential tool for distinguishing MRI signals of different temporal and structural patterns. References: 1. JF. Muzy et al,Int.Jour.Bif.,4,p.245,1994

S171

rithm which votes the results of many individual runs to recursively build a stable final partition [3]. Subjects and Methods: To assess the performance we compare the resulting true (TPs) and false positives (FPs) of the ensemble method with the individual clustering results using hybrid fMRI data sets constructed with functional CNR's of 1.33, 1.66 and 2 (h4h6). These hybrid data sets were constructed using a baseline in vivo MRI data set with artificially added activation (70 images, 64x64 pixels, 25 truely activated pixels). The basic cluster algorithm used was fuzzy c-means (FCM) [4] with a fuzzy index of 1.1, 20 initial number of clusters, random initialization, and Euclidean distance. The ensemble method is based on voting [3] the results of N=50 individual clustering runs. The resulting final set of clusters is obtaind by partitioning each pixel to the specific cluster where it has been assigned most often. For the activation cluster assessment of TPs and FPs was performed. Results: The activation cluster was easily determined as the one with the highest average voting result. The box-and-whisker plots in the figure show that the ensemble method achieves a much better definition of the truely activated pixels (indicated as *) for both TPs (left plot) and FPs (right plot) compared to the great dispersion of the 50 runs obtained by c-means alone. Discussion: The ensemble method based on many repetitions of an EDA method results in a final partition where (a) the activation cluster can be identified using the average voting result of the pixels in clusters and (b) which represents the true activation better than using individual results thereby reducing the instability induced by method and corresponding parameter choice. Acknowledgements: Supported by Austrian National Bank Fund (OeNB-Project-No.9201). References: 1. C.Windischberger et al, ArtIntelligMed, in press, 2003. 2. M.Barth et al, NMRBiomed,p.484-489(2001). 3. E.Dimitriadou et al, Int. J. of Pattern Recognition and Artificial Intelligence 16,p.901-912(2002). 4. J. Bezdek, Plenum Press(1981).

327 Ensemble clustering of fMRI data M. Barth1, E. Dimitriadou2, K. Hornik2, E. Moser3,1; 1Dept. of Radiodiagnostics, University and General Hospital, Vienna, AUSTRIA, 2Dept. of Statistics, University of Technology, Vienna, AUSTRIA, 3Dept. of Medical Physics, University, Vienna, AUSTRIA. Introduction: The main advantage of exploratory data analysis (EDA) approaches is that they do not require exact knowledge of the stimulation paradigm, of the individual hemodynamic response, or of the noise distribution, they rather search for various "interesting" time courses. Thus, EDA methods may be used to identify artifacts [1] and differentiate functional activation [2]. However, results may differ depending on method and corresponding parameters (initialization, metric, fuzzyness) used. To overcome this problem we used an ensemble clustering method based on an algo-

328 Time-domain quantitation of 1H short echo-time signals: background accommodation H. Ratiney1, Y. Coenradie2, S. Cavassila1, D. van Ormondt2, D. Graveron-Demilly1; 1Laboratoire RMN, CNRS UMR 5012, Université Claude Bernard, Lyon I-CPE, Villeurbanne, FRANCE, 2Applied Physics, Delft University of Technology, Delft, NETHERLANDS. Introduction: Quantitation of 1H short echo-time signals is often hampered by a background signal originating mainly from macromolecules and lipids. Several approaches have been proposed to handle the background signal. We compare three methods -- combined with the QUEST quantitation algorithm [1] -- in terms of bias-variance trade-off.

S172

Post-processing and Quantification: fMRI and MRS

Method: QUEST fits a combination of (quantum-mechanically simulated) signals of metabolites directly to the in vivo data at hand, in the time-domain. To get reliable quantitation results in presence of a background signal, the latter needs to be handled. Two approaches can be distinguished: 1. The background signal is included in the quantitation procedure. It can either be modelled with wavelets, splines [2] or polynomial functions or included in the basis set after measurement [3]. 2. The background signal is handled in a pre-processing step. Among these methods, we can cite in the time-domain 1) Truncation of initial data points which reduces its detrimental influence [4]; 2) SVD-based modelling and removal [5]. Existing methods -- except truncation -- suffer from strong correlation between the background and metabolite signals. We propose to truncate some initial data points, model by QUEST the resulting signal and subtract the background-related residue from the raw signal in a pre-processing step. Moreover, uncertainty caused by the background is accounted for by adding a correction term in the Cramér-Rao bounds (CRB). Results: Monte-Carlo studies were performed. For each noise level, 256 1H short echo-time signals of human brain at 1.5T were simulated, see Fig.1. They comprised twelve metabolites (aspartate, choline, GABA, glucose, glutamate, glutamine, lactate, myoinositol, N-acetylaspartate, phosphocreatine, creatine, taurine), lipids at 0.9 and 1.3 ppm, water and background signals modelled by five Lorentzian and three Gaussian components respectively. The signals were quantified with QUEST after water suppression with HLSVD and background handling. Three methods are compared in terms of bias-variance trade-off, see Fig.2.

Conclusion: QUEST, combined with the proposed pre-processing method, leads to estimates of the metabolite concentrations having the smallest bias and standard deviations. Moreover, the uncertainty of the background parameters has been accounted for in the Cramér-Rao bounds. Acknowledgements: This work is supported by Philips Medical Systems, Best, NL. References: 1. Ratiney et al., ProRISC-IEEE, 2002; 432-437. 2. Provencher, Magn Reson Med 1993; 30: 672-679. 3. Hofmann et al., Magn Reson Med 2002; 48, 440-453. 4. Stanley et al., Magn Reson Med 2001; 45: 390-396. 5. Y. Coenradie et al., MAGMA 2002; 15: 177-178.

329 Improved resolution metabolite maps from Multi-Echo Chemical Shift Imaging Data D. C. Williamson, N. A. Thacker, S. R. Williams; Isbe, University of Manchester, Manchester, UNITED KINGDOM. Introduction: The large voxels required in multi-echo chemical shift imaging (ME-CSI) contain signals from several different tissues which causes large variations of the metabolite signal from voxel to voxel1,2. This is especially true if a voxel contains cerebrospinal fluid (CSF). In order to identify regions of low metabolite concentration due to abnormalities, it becomes important to identify and correct for regions of low metabolite signal caused by normal tissue mixtures. Subjects and Methods: Spectroscopic data for a human brain were obtained for a single 2 cm slice with 32 x 32 pixels covering a 240 x240mm2 field of view using a 3 echo, ME-CSI sequence. In addition to this, reference images of 10 contiguous 2mm slices through the same volume are acquired using an inversion recovery acquisition. Data were acquired on a Philips ACS NT 1.5T scanner. The ME-CSI data are quantified using the j-MRUI3 implementation of the algorithm AMARES4. The results from the quantitation are exported from j-MRUI as ASCII files. The reference images are segmented using a Bayesian tissue classifier where the required probabilities are determined from the bivariate distribution of the grey levels and the grey-level slope5. This segmentation produces a 256x256-voxel map of the metabolite content per voxel. The map is reduced to a 32x32-voxel partialvolume map that is used to correct for CSF dilution in the metabolite data. Results and Conclusions: The combined data sets provide information about the quantities of metabolites in both grey and white matter. Figure 1 shows results for N-acetyl aspartate (NAA). The results for NAA, creatine and choline are in good agreement with the literature1,2. In addition the two data sets may be combined to produce metabolite images with the a resolution comparable to the reference images. Example results are presented in figure 2 (2a: reference image 2b: NAA map). Preliminary results suggest that differences in metabolites concentrations due to anatomical location are greater than differences between grey and white matter.

Post-processing and Quantification: fMRI and MRS This is in agreement with the findings of Schuff2. 1. M.A. McLean, et al. Magn Reson. Med. 44, 401–411, 2000, J. Pan. et al., Magn. Reson. Med., 40, 363-369, 1998., D. Weidermann et al. Magn Reson. Imag. 10, 1073-1080, 2001. 2. N. Schuff , et al. Magn Reson. Med. 45, 899-907, 2001. 3. A Naressi et al. MAGMA, 12, 141-152, 2001 4. S. Miersova et al. NMR Biomed 11, 32-39, 1998. 5. D. Williamson et al., MIUA2002, 17-20, 2002

330 A novel tissue segmentation method for metabolite quantitation from MRSI data P. Vermathen, C. Boesch, R. Kreis; Dept. Clinical Research, AMSM, University and Inselspital, Bern, SWITZERLAND. Introduction: For quantitative analysis of MRSI data, tissue segmentation has been used in several studies to correct metabolite values for CSF voxel contributions and to obtain metabolite concentrations in pure gray (GM) and white matter (WM). In this study we present a novel method for tissue segmentation, which employs multi-compartment fitting of water components in voxels at the resolution of the MRSI measurement. Methods: The method was tested in three subjects. Tissue segmentation: A modified Inversion Recovery Fast SpinEcho sequence was used to acquire 32 images with different individual echo times (17ms spacing). The sequence was repeated 5 times with TI =10-2010ms, yielding 180 intensity values with different combinations of TE and TI for each pixel. Images (FOV=16cm) were acquired with a 512×16 matrix, which was reduced afterwards to 32×16 and apodized, to match the MRSI resolution. Imaging time was ~5min. The data were fitted to obtain for each pixel M0 values for three compartments and thus the contribution of GM and WM, and CSF. T1 and T2 values were fixed for GM and WM and constrained for CSF. MRSI: A standard 2D-MRSI sequence with PRESS-volume preselection was used (TR/TE=1800/30ms, FOV=16cm, Matrix: 32×16). Postprocessing included moderate apodization, water

S173

removal and spectral fitting using the FITT software-package (courtesy of A.A. Maudsley). A separate water phantom scan served as external quantitation standard based on the reciprocity principle. Metabolite values were quantitated and corrected for CSF contributions. Results: Fig. 1 shows images of three compartments that were used for metabolite correction. Also shown are NAA images from the corresponding MRSI measurement. Regression analysis between CSF-corrected metabolite concentrations and WM voxel contributions (see Fig. 2) yielded for Cr in all three subjects and for Glu in two subjects significant differences between WM and GM. CSFcorrected metabolite concentrations in pure WM and GM were NAA: 10.4/10.6, Cr: 8.7/5.7, Ch: 2.1/1.8, Glu: 11.7/9.7 mmol/kg wet wt. in GM/WM, respectively.

Discussion: The presented method for tissue segmentation differs in several aspects from other ordinarily performed Methods: 1) generally a pixel is assigned exclusively to one compartment, while here image pixels can consist of all three compartments (though at coarser resolution); 2) since the matrices for imaging and MRSI were identical, the correction for point spread function is inherent ; 3) additional information is gained on water compartments. The method proved robust, metabolite values were similar to literature values. Acknowledgment: Swiss National Foundation (31-059082)

S174

Post-processing and Quantification: fMRI and MRS

331 Calculation of pure tissue spectra in proton MR spectroscopy of brain diseases U. Seeger1, T. Nägele1, I. Mader2, M. Erb2, U. Klose2; 1Dept. of Neuroradiology, University of Tübingen, Tübingen, GERMANY, 2Section Exp. MR of the CNS, University of Tübingen, Tübingen, GERMANY. Introduction: In single voxel proton MR spectroscopy of the human brain, it is often difficult to select volumes of interest (VOIs) that solely contain the tissue type of interest. In brain diseases where small or diffuse lesions have to be examined, considerable portions of normal brain are often included in the VOI and contaminate the spectrum. In this study, spectra of pure tissue types were calculated from short TE spectra of multiple VOIs with mixed tissue components. Subjects and Methods: Examinations were performed on a 1.5 T whole-body imager (Siemens “Sonata” or “Vision”) using either PRESS (TE 30 ms, TR 3,000 ms, 64 acquisitions) or STEAM (TE 15 ms, TR 3,000 ms, 64 acquisitions). Altogether 17 patients with lesions of various genesis were examined including 5 cases of chronic or acute encephalomyelitis, 5 glial brain tumors, 2 lymphomas, 3 meningiomas, 1 unspecific gliotic, and 1 ischemic lesion. A VOI of typically (2 cm)3 was placed in the area of the lesion. A further VOI was placed in a corresponding region of normal brain. The calculation of the pure tissue spectra from the measured spectra was based on the solution of a set of linear equations [1,2]. For this calculation, the tissue composition (percentage of the lesion, normal brain, and CSF) of each VOI was needed. It was determined by manual segmentation of additionally acquired MR images. Results: In most cases, the VOI of the lesion included considerable portions of normal brain. This contamination could be removed by calculation of the pure lesion spectrum. In Fig.1, an example is given showing a diffuse unspecific gliotic lesion only affecting white matter (the lesion appears bright in the T2-weighted images). In the measured spectrum of VOI 2 (including the lesion and normal brain), NAA is reduced, but still appearing. In contrast, the calculated spectrum of the pure lesion reveals a total absence of NAA and a strong increase of myo-inositole (mI) confirming an unspecific gliotic change. Discussion: The true metabolite content of a lesion without contamination from normal tissue is essential for the diagnostic information of the spectrum. The calculation of pure tissue spectra can improve the significance of localized spectroscopy.

Figure 1: For explanation see text.

References: 1. Wang Y, Li S-J [1998] Magn.Reson.Med. 39:28-33 2. Seeger U, et al. [2003] 11th Scientific Meeting ISMRM Toronto, in press Supported by DFG grants NA 395/1-1 and MA 2343/1-1.

332 Evaluation of two different methods for absolute quantitation of glycogen in 13C-MR spectra M. Ith, B. Jung, M. Zehnder, K. Zwygart, R. Kreis, C. Boesch; Department of Clinical Research, University and Inselspital Bern, Bern, SWITZERLAND. Introduction: Approaches for absolute quantitation of muscle glycogen, using 13C-MRS-resonance of C1, required calibration with anatomically shaped phantoms of a known glycogen solution [1], used creatine as internal standard [2], or evaluated sensitivityprofiles of surface-coils matched with MR-images [3]. Since the latter two methods were mainly used to determine signal-ratios or institutional units, this abstract shall expand and evaluate these methods for absolute quantitation. Subjects and Methods: 13C-spectra from thigh muscles of 18 trained volunteers (9 males/females) before and after a 3h trial on a bicycle ergometer [4] were acquired using a 1.5 T whole body MRscanner (GE SIGNA) with a double-tuned, flexible 13C/1H-coil (Medical Advance) applying a pulse and acquire sequence (adiabatic excitation; TR=301ms; CW-decoupling and NOE; 3x2000 acquisitions). An external acetone reference and two water filled tubes determined coil loading and position. Signal intensities S were determined by integration of peak areas and correcting for baseline-contributions. The creatine standard method uses total creatine as internal standard considering metabolite specific T1-saturation and NOE. Creatine concentrations of 32 mmol/kgww are used for our calculations in skeletal muscle. The image matched coil sensitivity map (csm) superimposes a calculated coil sensitivity profile to MR images with manually selected muscle tissue. Adding all csm-weighted selected volume elements leads to a theoretical signal αmuscle. Applying identical procedures to an arbitrarily shaped phantom filled with a known glycogen solution leads to αGlyc. Provided that both signals are individually corrected for coil loading the muscle's Glycogen concentration is calculated by multiplying the phantom concentration with αmuscle.Smuscle/ αGlyc.SGlyc. Results: The figure shows that strong correlations were found for males, females, and all volunteers. Confidence intervals for all interceptions include 0, however, slopes are significantly different from 1 in all groups.

Post-processing and Quantification: fMRI and MRS Discussion and Conclusion: Since the two methods showed strong correlation, both suit for relative quantitation [i.u.] but significant differences of absolute concentrations [e.g. mmol/kgww] indicate systematic under/overestimation for at least one of the methods. The creatine standard method overcomes problems generated by specific coil sensitivity profiles; however, it depends on the value of the creatine concentration and an accurate correction for T1-saturation and NOE. The image matched csm method avoids these problems by measuring a reference solution, however, it oversimplifies sensitivity profiles. These results are closer to generally accepted biopsy data than the creatine method. Acknowledgement: Supported by the Swiss National Fund (3100053788.98 & 3100-065315.01) References: 1. Gruetter R, Magnusson I, Rothman DL, Avison MJ, Shulman RG, Shulman GI [1994] Magn Reson Med 31:583. 2. Jehenson P, Duboc D, Bloch G, Fardeau M, Syrota A [1991] Neuromuscul Disord 1:99. 3. Slotboom J, Flück C, Kreis R, Jung B, Nuoffer JM, Mullis P, Boesch C [1998] Proc ISMRM 6:1860. 4. Zehnder M, Ith M, Jung B, Kreis R, Saris W, Boutellier U, Boesch C [2002] Proc ISMRM 10:167. 5. Bottomley PA, Lee Y, Weiss RG [1997] Radiology 204:403.

S175

required, provided the frequencies are well separated and the line broadening is identical across the spectrum. However for adjacent frequencies a correlation develops between the amplitudes as the peaks broaden, invalidating the use of either as a reference (Figure 1).

333 Assessment of the Pade approximant as a method for quantifying in vivo proton Magnetic Resonance Spectroscopy data D. C. Williamson, N. A. Thacker, S. R. Williams; Isbe, University of Manchester, Manchester, UNITED KINGDOM. Introduction: Recently Belkic1 presented the Padé approximant (PA) as a method of generating magnetic resonance spectra with superior resolution to the discrete Fourier transform. Here we assess the PA as a parametric estimator and compare it with results from HLSVD2 and AMARES3 using simulated and experimental data. Methods/Subjects: Simulated signals were generated with peaks to represent, lactate (doublet), N-acetylaspartate, creatine (2 peaks) and choline. Signals with varying lineshape (Lorentzian, Voigt) were simulated in the presence of 10% white noise. In addition, single voxel MRS data were acquired from a seven-tube phantom. Each tube contained a different concentration of acetate, creatine and choline. The data were quantified using the PA, implemented according to the algorithm described by Belkic1. The results were compared with those obtained using HLSVD and AMARES as implemented in jMRUI5. Results: Table 1 shows the N-acetylaspartate amplitudes obtained by quantifying the simulated data. For Lorentzian lineshapes, all three methods are identical. As the Gaussian character of the Voigt lineshape is increased all methods show a similar bias in the mean. Table 1: NAA amplitudes obtained using PA, HLSVD and AMARES to quantify simulated signals. Gaussian Linewidth Noiseless Amplitude PADE (Hz.)/Pi (STD)

HLSVD (STD)

AMARES (STD)

0

25

24.98 (0.46) 24.99 (0.49) 25.00 (0.42)

4

25

26.42 (0.44) 26.78 (0.41) 26.96 (0.41)

8

25

28.78 (0.59) 29.03 (0.60) 29.14 (0.55)

12

25

29.13 (0.76) 31.43 (2.46) 30.20 (0.70)

Using a Lorentzian lineshape in the model results in an overestimation of the amplitude. This is not a problem if amplitude ratios are

Figure 2 shows the ratio of creatine and acetate amplitudes determined from the quantification of the phantom data versus the equivalent ratios of concentrations. Similar results were found for choline and for acetate, where the amplitude in tube 1 is used as a reference for the other tubes.

S176

Animal Models: Imaging and Microscopy

Conclusions: The PA algorithm performs as well as the other algorithms with respect to quantifying spectroscopy data. The PA has the advantage that the amplitudes can be calculated analytically without the need for fitting or determining the number of peaks. These results suggest that in some instances the Lorentzian assumption may be inadequate and the use of an alternative lineshape is required. Both the PA and AMARES could be generalized, albeit non-trivially, whereas HLSVD requires the Lorentzian lineshape assumption. 1. Dž. Belkic et al. J. Chem Phys., 113, 6543-6556, 2000. 2. W. Pijnappel et al. J. Magn. Reson. 97, 122-134,1992. 3. S. Miersova et al. NMR Biomed 11, 32-39, 1998. 4. A Naressi et al. MAGMA, 12, 141-152, 2001

Scientific Session Animal Models: Imaging and Microscopy 11:20 am - 1:00 pm

tion of the vessel wall (for instance in case of atheroma, the volume, localization and distribution of the lipid core, fibrous tissue and calcifications, the thickness of the fibrous cap...) become accessible non invasively and using a standard scanner. High-resolution MRI using this HTS probe offers the potential to study in vivo superficial vessels and pathologic vascular processes such as athero-thrombosis or inflammatory diseases. References: 1. Falk E. et al. [1995]Circulation.92:657-71. 2. Toussaint JF. et al. [1996] Circulation.1994:932-8 3. Ginefri J-C. et al. [2001] Mag. Res. in Med. 45:376-382

Ruys

334 Ex-vivo 3D microscopy of coronary arteries using a standard 1.5T MRI scanner and a HTS RF probe M. L. Poirier-Quinot1, J. Ginefri1, F. Ledru2, P. Fornes3, L. Darrasse1; 1U2r2m, CNRS/UPS UMR8081, Orsay, FRANCE, 2Cardiology Department, Hospital European Georges Pompidou, Paris, FRANCE, 3Pathology Department, Hospital European Georges Pompidou, Paris, FRANCE. Purpose: A major issue in current cardiovascular research is to develop imaging technologies capable of identifying unstable atheromatous plaques in vivo 1. Although MRI is a good candidate for this purpose2, an isotropic spatial resolution under 100µm would be needed to fully characterize the different components of the plaques. Although current clinical scanner with typical field strength of 1.5T have gradients strong enough to reach such a limit, a much-too-low SNR problem then arises. We have investigated on excised (but non dissected) human hearts the improvement achieved by calling to a high temperature superconducting (HTS) RF probe dedicated to localized microscopy in humans3. Subjects and Methods: Our superconducting surface probe, with effective diameter of 12 mm, has an unloaded quality factor outside/inside the magnet of about 60000/11000. Two hearts were collected from patients soon after death and kept in formaldehyde. 3D data sets were acquired within 25min (T1-weighted Gradient-echo sequence, FOV=30x15x12.4mm3, matrix=512x256x124). Extended exploration of the coronary artery was achieved using a graduated plastic support to translate the heart along the HTS coil. Results: MRI cross sections of the left anterior descending artery (Fig 3 and 5) are compared to the corresponding histological cross sections (Fig 2 and 4 staining at 40 x magnification). The morphology, structure and components of the plaques are well recognized and delineated. Fig.1 shows a view reconstructed along the artery axis from 2 data sets. Discussion: We have demonstrated that a HTS can achieve highly resolved details within the excised heart, which can be confronted with histological slices. Contrarily to histology, MRI can provide extended 3D information without deformation. Geometry of the arterial lumen (percent area obstruction, eccentricity, abruptness of the inlet or outlet...), as well as the geometry and tissue composi-

335 Feasibility of unicellular stem cells imaging in rodents brain B. Gillet1, B. Doan1, E. Pottier2, C. Riviere3, H. Petite2, F. Gazeau3, J. Beloeil1, P. Meric1; 1Icsn, CNRS, Gif sur Yvette, FRANCE, 2Lro, CNRS, Paris, FRANCE, 3Lmdh, CNRS, Paris, FRANCE. Introduction: A recent study have succeeded to show by MRI in the rat, the massive migration of embryonic stem cells loaded with ultra-small superparamagnetic iron oxide particles from the cerebral implantation site toward an ischemic lesion in the contra lateral hemisphere, however the sensitivity of the MRI technique was limited to the detection of clusters of at least 40 cells (1). Our purpose was to develop MRI technique with a sufficient spatial and temporal resolution to allow in vivo longitudinal studies of the mapping of implanted stem cells up to the individual level. Subjects and Methods: Cell Labeling: Stem cells were obtained from in vitro-expanded marrow stromal cells C57Bl6 mice (2). Labelling of the cells was obtained by incubation in culture media containing a new type of magnetic iron-oxide nanoparticles (3). Negative charges at the surface of these nanoparticles allow their spontaneous incorporation in the cells by endocytose.The efficiency of the cell labelings were controled by measurement of their iron contents by magnetophoresis and RPE. MRI: Experiments were performed on a vertical wide bore 9.4T VARIAN Inova equipped with actively shielded gradients sets of 2T/m (RSI). In vitro acquisitions were obtained from a loop-gap coil (l= 12 mm Ø= 8 mm) with 2D or 3D experiments. In vivo spin echo images of the mouse brain were obtained from surface coil (Ø= 10mm). Results: Magnetophoresis and RPE show that labeling of cells was proportional to the concentration of nanoparticles in the incubation medium in the range of 1 to 5mM and that labeling remained de-

Animal Models: Imaging and Microscopy tectable up to 14 days of culture after labeling with 1mM. 2D imaging of cells in agarose gel(70 ¢/µl) show no significant effect on T1 and T2 of the mixture. 3D T2*Imaging (voxel resolution 30µ x 30µ x 70µ ,TR/TE 200 ms/2 ms, acq. time 120 min.) show black spots of about 80µ in diameter with shapes and concentration corresponding to those of the cells in agarose gel (figure). In vivo brain mouse imaging with spatial and temporal show suitable signal to noise ratio for similar detection. Conclusion: The feasibility of in vivo detection of individual implanted cells is strongly suggested by the results. Further studies have to determine the feasibility and the potential duration of longitudinal investigations. References: 1. Hoehn M et al. [2002] PNAS 99:16267-16272 2. Petite H et al. [2000] Nat Biotechnol. Sep;18:929-30 3. Wilhelm C et al. [2002] Langmuir 18 : 8148-8155.

S177

frame were computed from separate acquisitions of motion encoding in read-, phase-, and slice-directions. The acquisition was ECG and respiratory gated using a home-build ECG unit [2]. 15 frames of midventricular short axis and long axis views were obtained over the RR-interval (130ms) to cover a heart cycle. To acquire all information for computing the 3D velocity, 7 cine data sets were recorded in about 30min. Results: The measuring methods were validated on a rotating and a tube-phantom yielding deviations of only 1.5 %. Figure 1 and 2 show one frame of the midventricular myocardium short axis and long axis, respectively. For comparison, the pictures were recorded at the same position of the heart cycle. Each arrow represents the velocity magnitude and direction in the corresponding image voxel (size: 78x78x500µm³) with a temporal resolution of about 5 ms. We were able to measure absolute myocardial velocities ranging from 0.5 to 3.5cm/s over the heart cycle. These velocities are higher than planar motion measurements [3] reflecting the contribution of the third dimension.

336 Quantitative 3D in vivo motion mapping of murine myocardium with phase contrast MRI at 17.6 Tesla P. Mörchel, V. Herold, C. Faber, E. Rommel, A. Haase; Experimentelle Physik V, Universität Würzburg, Würzburg, GERMANY. Purpose/Introduction: Characterization of myocardial movement delivers a direct measure of the function and viability of the heart. The aim of this work was to assess high temporal and spatial resolution cine-sequences with 3D phase contrast motion encoding [1] of the murine myocardium on a 17.6T MR-system. Subjects and Methods: All measurements were performed on a Bruker Avance 750 spectrometer with a maximum gradient strength of 1.0T/m and a 20mm Bruker birdcage coil. Mice with body weights between 10 and 20g were anesthetized via 1.0 – 1.5 vol.% isoflurane inhalation. Images were acquired using a segmented FLASH sequence with velocity compensation in all gradient directions (TE 2.7ms, TR 8.0ms, resolution 78x78µm, slice thickness 0.5mm, 2 averages). Motion encoding was done by preparing the spin phase using a bipolar gradient which causes the moving spins to accumulate a velocity-dependent net phase with respect to stationary spins. The 3D velocity vectors of each cine

Discussion/Conclusion: High field phase contrast MRI provides high resolution motion maps. We have demonstrated that this method delivers accurate quantification of the velocity in read-, phase- and slice-directions at a spatial resolution of 78µm in plane and a temporal resolution of about 5ms. References: 1. Van Dijk P [1984] JCAT, 8(3):429-436 2. Rommel E et al [2000] Proc ESMRMB, Paris, 568 3. Streif J et al [2003] MRM 49:315-321

S178

Animal Models: Imaging and Microscopy

337

338

Transplanted embryonic stem cells labeled by superparamagnetic nanoparticles monitored by MRI V. Herynek1, K. Glogarova2,3, J. Kroupova2,3, P. Jendelova2,3, L. Urdzikova2,3, B. Andersson4,3, P. Dvorak2,3, M. Hajek1, E. Sykova2,3; 1MR-Unit, Dept. Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague 4, CZECH REPUBLIC, 2Center for Cell Therapy and Tissue Repair, 2nd Medical faculty, Charles University, Prague 5, CZECH REPUBLIC, 3Dept. of Neuroscience, Institute of Experimental Medicine ASCR, Prague 4, CZECH REPUBLIC, 4Dept. of Clinical Neuroscience, Karolinska Institutet, Stockholm, SWEDEN.

Improved High Resolution 23Na Imaging of the mouse heart T. Neuberger, A. Greiser, M. Nahrendorf, A. Haase, C. Faber, A. Webb; Physikalisches Institut, EP5, Universitaet Wuerzburg, Wuerzburg, GERMANY.

Introduction: Cell transplantation is potentially a powerful treatment method. Pluripotent cells could replace damaged tissue with low recovery ability. We have used a superparamagnetic contrast agent to visualize the fate of non-differentiated (ND) and partially differentiated (PD) embryonic stem cells (ESC) transplanted into adult Wistar rats with an unilateral cortical photochemical lesion in vivo using MR imaging. Subjects and Methods: The commercially available mouse embryonic stem cell line D3 was stably transfected with pEGFP-C1 and co-labeled with superparamagnetic iron-oxide nanoparticles (Endorem™, Guerbert Laboratories, France) 3 days prior to transplantation. Transmission electron microscopy confirmed the presence of iron-oxide nanoparticles inside the cells. Neural differentiation was induced 5 to 8 days prior to transplantation. The cortical lesion was induced by Bengal rose/light beam interaction 7 days prior to transplantation. ND or PD cells were administered either into the contralateral hemisphere (30.000-250.000 cells in 5 µl PBS; ND: n=6 animals; PD: n=10) or intravenously into the femoral vein (300,000-1.600,000 cells in 0.5 ml PBS; ND: n=3; PD: n=6). T2W images (turbo-spin echo sequence, TE/TR= 42.5/2000 ms) were taken weekly using a 4.7 Bruker MR spectrometer. Rats were sacrificed 4 weeks following transplantation, and the fate of the transplanted cells was analyzed immunohistochemically. Results: The implanted cells (in the case of intracerebral administration) were visible on MR images as a sharply bounded hypointense area at the injection site. Starting 7 days after transplantation and persisting for 4 weeks, a hypointense signal was detected in the lesion (see Figure). Hypointensity was also found in the lesion after the i.v. injection of labeled cells. Immunohistochemistry subsequently confirmed the presence of iron-rich cells inside the lesion. Discussion/Conclusion: Our study demonstrates that brain grafted and intravenously injected ESCs migrate into a lesion site. Partial differentiation of the cells prior to transplantation led to their increased presence in the lesion. The commercially available contrast agent Endorem can be used as a marker for long-term noninvasive MR tracking of implanted ESCs in the brain. Supported by: LN 00A065, J13/98: 11130004, AVOZ5039906I024, GACR 304/03/1189, CEZ:L17/98:00023001 Figure: An animal with contalaterally (A) or intravenously (B) injected Endorem-labeled ESCs, two weeks after cell implantation.

Introduction: Sodium concentration plays an important role in cardiac metabolism. Early detection of myocardial infarcted areas can potentially be achieved by observing the increase in concentration in the newly-created extracellular space due to cell death. In this work we present results from non-invasive in vivo proton and sodium magnetic resonance cardiac imaging. Significant improvements have been made in terms of spatial resolution and signal-tonoise (S/N) compared to earlier studies [1]. Subjects and Methods: Experiments were performed on a Bruker 17.6 T widebore Avance System with a maximum gradient strength of 200 mT/m. After positioning the anesthetized mouse in the magnet, CINE-imaging [2] was performed (12 frames/heartbeat; res=0.3x0.3x0.8 mm; NS=8 ; Ttot=6 min ). In addition to visualizing cardiac dynamics, these images enable subsequent spatial correlation of proton and sodium images. Sodium images were obtained using a pure phase-encoded, density-weighted 3D localization method [3] in which the k-space sampling scheme is designed to provide an optimized shape of the spatial response function (SRF), minimal Texp and high SNR. With a non-selective short excitation pulse (80 µs) and an adapted gradient waveform the pre- acquisition delay was minimized (< 0.75 ms) to provide maximum SNR. The spatial resolution corresponds to an isotropic 1.1 µl voxel size. Due to the requirements for respiratory and ECG triggering, sodium imaging took about 2 hours (4 scans/heartbeat; 130,000 total scans). Results: Figure 1 shows a short and a long axis cut through a mouse heart 4 weeks after myocardial infarction (anterior left ventricle wall infarct) from the CINE data sets. The ventricles and the myocardium are shown with a good contrast to noise ratio. Four short axis cuts from the 3D 23Na density-weighted dataset of the same mouse are shown in figure 2. The SNR of the septum is about 60% less than in the ventricles. The arrows indicate the location of the infarcted area. Discussion: Sodium microimaging at high SNR (>15:1) and high spatial resolution (~1 µl) shows promise for the detection of small myocardial infarcts in a mouse model. In combination with highresolution proton CINE–images both structure and function in such models can be investigated. References: 1. Greiser A et al [2002] Proc. ESMRMB: 64 2. Rudin M et al [1991] Basic Res Cardiol 86(2):165-74 3. Greiser A et al [2001] Proc. ISMRM 9: 718.

Animal Models: Imaging and Microscopy

S179

339

340

Embryonal and fetal brain development in bovines, a MR-Histology study D. Haddad1, A. Purea1, M. Schmidt2, M. Haas-Rioth2, H. H. A. Oelschläger2, A. Haase1; 1Experimentelle Physik V, Physikalisches Institut, Universität Würzburg, Würzburg, GERMANY, 2Dr. Senckenbergische Anatomie (III), Johann Wolfgang Goethe-Universität, Frankfurt a.M., GERMANY.

MRI of transplanted pancreatic islets. A preliminary in vitro study D. Jirák1,2, J. Kríž3, H. Vít1, B. Andersson4,5, P. Girman3, M. Burian1,2, F. Saudek3, M. Hájek1,2; 1MR-unit, Dept. Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague 4, CZECH REPUBLIC, 2Center for Cell Therapy and Tissue Repair, 2nd Medical faculty, Charles University, Prague 5, CZECH REPUBLIC, 3Laboratory of Islets of Langerhans, Institute for Clinical and Experimental Medicine, Prague 4, CZECH REPUBLIC, 4Dept. of Clinical Neuroscience, Karolinska Institutet, Stockholm, SWEDEN, 5Dept. of Neuroscience, Institute of Experimental Medicine ASCR, Prague 4, CZECH REPUBLIC.

Introduction: In 'soft-tissue studies' MR-Histology is a fast and efficient alternative to conventional histological sections, since in this type of tissue MR methods are able to provide a distinct and adjustable contrast and lead to the desired information without the tedious preparation process which is necessary for conventional histological sections. Here we present first results of an ongoing study on the development of the prenatal bovine brain. The study is performed, because neither zoological nor veterinary anatomical textbooks provide detailed information about bovine brain development and therefore comparative studies with other mammals are still impossible. Especially allometric growth and differences in the development of the brain are key issues in this context. Seven formalin fixed specimens at different stages in the prenatal bovine development were chosen. Their crown-rump-lengths (CRL) range from 1.8-8.0cm, thus covering a large period in the prenatal bovine development. The smaller embryos still show the common mammalian organization, whereas the larger fetuses already exhibit typical bovine features like rudiments of the hooves. On account of the high water content, preparation of embryonic brain tissue in toto without severe damage is almost unfeasible. MR-Histology leaves the embryos uninjured, and provides life-like renderings of the natural structures. Materials and Methods: So far, the smaller specimens with CRLs up to 2.9cm were scanned in a 20mm birdcage coil on a Bruker Avance-750 spectrometer (17.6T). Different MRI sequences were used, depending on the desired contrast and spatial resolution of the images. Isotropic spatial resolutions of (58.6µm)³ were easily reached. During data processing a zero-filling by a factor of 2 was applied in each direction.

Results and Discussion: The MR images of the already scanned specimens exhibit various details of the prenatal bovine anatomy. In the brain, different substructures are distinguishable (cf. Fig.1). The eyes, the tongue and the nose with the air passage as well as the larynx are also clearly visible. Furthermore, the upper and lower jaws and the spine are to be seen. In conclusion, MR-Histology was able to provide the desired anatomical details to visualize the prenatal brain development in bovines. After scanning the remaining specimens, the images will be segmented to perform surface reconstructions of the different brain compartments. With these reconstructions, the development of the bovine brain can be displayed three-dimensionally and the volumes and surface areas of the brain's subcompartments can be calculated. This allows a quantitative comparison with the brain development in other mammals.

Introduction: Transplantation of pancreatic islets (PI) is a promising treatment method of Type-1 diabetes mellitus. The most frequent place of PI injection is portal vein. Up to now there is no available method which can provide an information about number and distribution of PI in the liver in vivo. The aim of our in vitro study is to consider possibility of PI visualization after transplantation into the liver using magnetic resonance imaging (MRI). Subjects and Methods: PI were harvested from adult male Wistar or Lewis rats and 2 days cultured in medium CMRL-1066 (37°C, 5% atm. CO2; Sigma) with a superparamagnetic contrast agent Resovist (100µl/10ml (n=8) or 50µl/10ml (n=5); Schering). PI were transplanted into the liver through the portal vein (manual injection of 2500 islets) of the adult male Wistar (n=8) or Lewis (n=5) rats respectively. In case of the Wistar rats, the portal vein was ligated 5 minutes after transplantation, the liver was removed and put into saline. In case of the Lewis rats, the liver removal was performed after one week. MR images (gradient echo sequence, TE/TR= 3.4/80 ms) were taken within two hours after liver removal using a 4.7 Bruker MR spectrometer equipped with a homemade surface coil. Two liver samples were analyzed by transmission electron microscopy (TEM). Results: The labeled PI were visible on MRI as a sharply bounded evenly distributed hypointense spots in the liver immediately after transplantation and also one week after transplantation (see Figure). TEM confirmed presence of iron-oxide nanoparticles inside the cells (mitochondrion).

Figure: Islets of Langerhans transplantation into the liver, in vitro MRI was performed one week after transplantation. Discussion/Conclusion: The used superparamagnetic contrast agent visualize the PI in the liver tissue in the MR images. Our study demonstrates that transplanted PI are visible in the liver also

S180

Perfusion

one week after PI transplantation. Promising results of this preliminary study represent a significant step toward monitoring of the fate of transplanted PI in vivo. Supported by: VZ/CEZ: L17/98:00023001, LN00A065

The authors acknowledge the Biological Imaging Centre, Imperial College London for MR expert advice and use of MR facilities.

341 Magnetic Resonance Imaging of inflammation - Targeting leucocyte infiltration P. R. Reynolds1, D. J. Larkman2, J. V. Hajnal2, D. O. Haskard3, A. J. T. George4, A. D. Edwards5; 1Immunology and Paediatrics, Imperial College London, London, UNITED KINGDOM, 2Imaging Sciences Department, Imperial College London, London, UNITED KINGDOM, 3BHF Cardiovascular Medicine, Imperial College London, London, UNITED KINGDOM, 4Immunology, Imperial College London, London, UNITED KINGDOM, 5Paediatrics, Imperial College London, London, UNITED KINGDOM. Purpose: Migration and infiltration of inflammatory cells is an underlying mechanism in all inflammatory diseases. We are developing core technologies to image inflammatory lesions by magnetic resonance (MR) using superparamagnetic contrast agents. Ultrasmall iron oxide nanoparticles (USPIO) are iron-based contrast agents that accumulate in phagocytic cells and cause considerable T2 shortening (1). This study addressed two questions: 1) Is the time course of T2-shortening in an inflammatory lesion after USPIO injection consistent with contrast uptake by trafficking phagocytic cells? 2) Where does the contrast-labelling of phagocytes occur ? Methods: Oxazolone-induced contact hypersensitivity was induced in the ears of 4-6 week old BALB/c mice (2). 18-24 hours after the challenge ear thickness was measured, and baseline MR imaging of the ear performed on a 9.4-Tesla Varian horizontal-bore MR scanner using a custom-built ear coil. A T2 measurement protocol was developed using spin-echo sequences (TE5.5, 15, 25, 50, TR 1500). USPIO were administered intravenously after baseline imaging. Imaging data was fitted to a T2 decay model; goodness of fit, intercept and T2 parametric maps were generated. Regions of interest were analysed, and T2 values plotted. This was repeated over 28 days post-injection to produce a time course of T2 change. Leucocyte purification with imaging and staining was used to determine the process of labeling. Histology was performed. Results: Maximal swelling was observed 18-24 hours post-challenge. In the ears of mice given USPIO there was significant T2 shortening, which persisted after clearance from the vascular compartment, whilst control T2 values were not significantly altered (Figure 1). Maximal shortening occurred around day 4. Cell infiltration was confirmed on histology (Figure 2). Contrast-labelling of phagocytes appears to take place in the extravascular compartment, not in the vascular compartment. Conclusion: USPIO are taken up by phagocytes in-vivo and can be imaged using MR. Change in T2 measurement provides a robust method for serial studies of this type as it is independent of coil/animal registration and has the potential to provide concentration information. Our results are consistent with extravascular labelling of phagocytic cells. Further studies are needed to define the precise mechanism of uptake. Cell specific imaging of inflammatory lesions has the potential to be an important investigative tool for inflammatory diseases. References: 1. Jung CW, Jacobs P. Magn Reson Imaging;13(5):661-74 2. Harari OA et al J Immunol;163(12):6860-6

Scientific Session Perfusion 11:20 am - 1:00 pm

Plate

342 Assessment of quantitative contrast agent concentration in dynamic magnetic resonance First-Pass perfusion imaging F. Fidler, P. M. Jakob, A. Haase; Physikalisches Institut EP5, Universität Würzburg, Würzburg, GERMANY. Introduction: First-pass perfusion imaging was shown to be a robust and reliable MRI method for qualitative analysis of altered perfusion in human myocardium and for quantification of perfusion reserve. Commonly it is optimized to assure that the measured signal intensity reflects tracer concentration by uses of a low CA dose. Purpose of this work was to exploit a model for quantitative CA concentration evaluation from first-pass data to overcome this assumption and to allow for quantitative analysis of perfusion with increased CNR by using higher CA dose. Methods: CA concentration ccalc(s) is derived from signal intensity s(cbolus) from T1-weighted imaging experiment is described in our approach phenomenological by ccalc(s) = ( [ exp{s(cbolus)*p/s(cbolus=0)} * exp{-p} ] – 1) * h with p = 1 / (b*T1- a) and (a) Parameters a, b, h depend on measurement parameters derived

Perfusion from simulation or an additional calibration measurement (b) T1 has to be measured before each CA experiment which yield in T1 of blood and tissue (c) Baseline intensity s(cbolus=0) is the apparent signal intensity of precontrast images Signal output from a first-pass saturation prepared FLASH with TI=10ms, TR=2.9ms, TE=1.1ms, FA=18°, T2=50ms, T1blood=1700ms, T1heart=1350ms was simulated (fig.1) and the exponential model was applied. Tissue response to a given gammavariate shaped arterial input function (fig.2) was calculated by convolution with a fermi function. Model parameters were optimized for given method to a= 0.30, b=0.0082, h=1.84. Model was applied to simulated signal intensities to extract concentration time curves (fig.2). As a representative deconvolution model we calculated mean blood flow with fermi deconvolution [1] for a wide range of bolus concentrations (fig.3). Results: The new model improves linear dependency to tracer concentration (fig.1) and calculates CA concentrations. With this model perfusion evaluation accuracy and deconvolution is improved (fig.3). This implies the possibility to increase CNR by using a 3-fold higher CA dosage. This model is applicable to a wide range of methods like saturation prepared SSFP. Conclusion: The proposed exponential model yield a phenomenological correction for quantitative perfusion evaluation in the high CA dose regime with an increased CNR. 1. M Jerosch-Herold et al., Med.Phys.25(1),1998

Figure1: Measured CA concentration

Figure2: Simulated high CA dose first-pass experiment using quantitative exponential model (solid line), linear model (dashed line) and given concentration (dotted)

S181

Figure3: Mean blood flow evaluation from deconvolution with fermifunction using quantitative exponential model (solid line), linear model (dashed) and given MBF (dotted)

343 Pre bolus technique for MR quantification of myocardial perfusion H. Köstler, C. Ritter, M. Lipp, M. Beer, D. Hahn, J. Sandstede; Institut für Röntgendiagnostik, Universität Würzburg, Würzburg, GERMANY. Purpose: The dose of contrast medium (CM) for analysis of first pass myocardial perfusion depends on the method of data post processing. A higher dose of CM optimizes the visual data evaluation due to a better signal-to-noise ratio (SNR), but a quantitative analysis of high dose studies is not possible because saturation effects in the ventricle deny the determination of the arterial input function (AIF). Therefore, a combination of a low dose for the determination of the AIF with a higher dose for an improved SNR (pre bolus) would be useful. Subjects and Methods: 18 first-pass perfusion studies in healthy volunteers were performed using a 1.5 T MR-scanner with multi slice truefisp perfusion sequence (TR/TE/TI 2.6/1.1/110 ms; FA 50°, SL 8 mm, acquisition-time 192 ms, flow rate 4 ml/s, 40 images at one image per heart beat). The pre bolus technique combines two first-pass series with two consecutive injections of a smaller and a larger bolus within 1 min. We compared the following protocols: (I) 3 ml, 9 ml, 12 ml Gd-DTPA with a delay of at least 15 min between the injections; (II) pre bolus-technique 1 ml / 8 ml vs. 1 ml / 12 ml Gd-DTPA; (III) pre bolus 1 ml / 8 ml Gd-DTPA vs. 1 ml / 4 ml Gadobutrol. We evaluated the SNR and perfusion values after deconvolution with the AIF. Results: (I) SNR increased with a rising CM-dose. The evaluation of perfusion parameters after deconvolution with the AIF was only possible if the AIF was taken from the 3 ml series. (II) The pre bolus technique allowed a deconvolution with the AIF and provided an improved SNR. There was no significant difference of the SNR with 1 ml / 8 ml and 1 ml / 12 ml. (III) Perfusion values and SNR showed no significant difference after deconvolution for 1 ml / 8 ml Gd-DTPA vs. 1 ml / 4 ml Gadobutrol. Conclusion: The pre bolus technique allows the determination of the AIF and the myocardial concentration time courses using two consecutive contrast agent injections. This enables semi-quantitative evaluation of myocardial perfusion with an improved SNR.

S182

Perfusion

344 Quantitative Myocardial Perfusion Imaging using ContrastEnhanced Dynamic MRI W. G. Schreiber1, M. Schmitt1, S. E. Petersen2, G. Horstick2, A. Karg1, N. Hoffmann1, K. F. Kreitner1, J. Meyer2, M. Thelen1; 1Dept. of Radiology, Mainz University Clinics, Mainz, GERMANY, 2Dept. of Cardiology, Mainz University Clinics, Mainz, GERMANY. Introduction: Contrast enhanced T1 weighted magnetic resonance imaging (MRI) is an interesting non invasive method for assessment of coronary artery disease. However, to date quantification of myocardial blood flow (MBF) with ultrafast MRI has not reached the same acceptance as MBF quantification with PET. It was the purpose of this study develop a methodology to assess MBF from dynamic contrast enhanced MRI, validate it with a gold standard and to test its applicability in a pilot patient study. Subjects and Methods: Dynamic contrast-enhanced MRI was performed before, during and after a single bolus injection of 0.025mmol/kg body-weight Gd-DTPA using a saturation-recovery snapshot fast low angle shot pulse sequence. Determination of MBF was performed using nonlinear data fitting based on a mathematical model of myocardial tissue (MMID4). Validation was performed in a pig model of acute myocardial ischemia (n=6) using fluorescent microspheres as a gold standard. In a patient study in seven patients with coronary artery disease, the myocardial perfusion reserve was determined based on two MBF determinations, one under resting conditions and one during hyperemia induced by Adenosine. Results: The study showed a correlation of R=0.75 for MBF quantification with contrast enhanced MRI and fluorescent microspheres. Within error limits, MBF was neither over- nor underestimated in ischemic (c.f., Fig. 1) and non ischemic (c.f., Fig. 2) myocardium, respectively. In the patient study, resting MBF values in patients were 0.8±0.2mL/min/100g and 1.0±0.2mL/min/100g in non-ischemic and in ischemic regions, respectively. Adenosine induced a significant increase of the MBF in all but one patient. The corresponding values were 1.4±0.7mL/min/100g and 2.5±1.0mL/min/100g in non-ischemic and in ischemic regions, respectively. The myocardial perfusion reserve was significantly reduced in ischemic regions when compared with non-ischemic regions (2.0±1.0 vs. 3.7±2.3, p<0.001). Discussion and Conclusion: In animals, a good correlation between MRI derived MBF estimates and those from fluorescent microspheres was observed. Moreover, the results of the MBF determination in patients correlate nicely with published values in the literature (MBF = 1.0±0.3mL/min/100g under resting conditions; Hyperemia: 2.1±1.2mL/min/100g and 3.4±1.3mL/min/100g in non-ischemic and in ischemic myocardium). Our results demonstrate that MBF can be determined reliably using contrast-enhanced dynamic magnetic resonance imaging.

345 Influence of smoking on pulmonary function in humans using a combined 1H MR O2-enhanced T1 mapping and spin labeling protocal T. Wang1, A. Rauch2, G. Schultz3, D. Hahn3, P. M. Jakob1; 1Department of Physics, University of Würzburg, Würzburg, GERMANY, 2Department of Cardiology, University of Würzburg, Würzburg, GERMANY, 3Department of Radiology, University of Würzburg, Würzburg, GERMANY. Purpose/Introduction: Recently, the influence of smoking on pulmonary perfusion in humans using 1H MR spin labeling technique has been investigated and the preliminary results revealed that the pulmonary perfusion increased after smoking (1). In this work, we further compare pulmonary function before and after smoking by measuring arterial blood and lung T1 using combined 1H MR O2enhanced T1 mapping (2) and spin labeling technique (3) because it is well-known that the oxygen partial pressure, pO2, in arterial blood provides a general assessment for pulmonary function. Subjects and Methods: Two “healthy” smokers were examined in the supine position on 1.5-T VISION, Siemens. Informed consent was obtained before each study. The smokers were asked to prevent smoking for one day before the examination. T1 mapping was performed in heart and lung in combination with breathhold on endexpiration, while the subjects were breathing 21% O2 and 100% O2, respectively. Additionally while breathing 21% O2, pulmonary perfusion rates were also calculated using spin labeling technique (1,3). No ECG triggering was used in all studies. After a break of about 20 minutes accompanied with smoking 2–4 cigarettes, the measurements were repeated. Results: An acquired representative T1 map from a 23-year-old male smoker while breathing 21% O2 was shown in Figure 1 and ROIs including 30 pixels were drawn in the left ventricle and the lung. The corresponding R1 (mean ± SD) while breathing 21% O2 and 100% O2 was plotted as a function of inhaled O2 concentrations before and after smoking, see Figure 2. Before smoking, arterial blood and lung show ∆R1 slopes of O2 enhancement of 0.17 and 0.10 1/s/%O2, respectively. After smoking, arterial blood and lung show the reduced ∆R1 slopes of O2 enhancement of 0.12 and 0.06 1/s/%O2. Note also in Figure 2 that both arterial blood and lung T1 after smoking are reduced while breathing 21% O2. According to spin labeling experiments, the lung shows a pulmonary perfusion rate of 467 ml/100g/min before smoking and 599 ml/100g/min after smoking. Discussion/Conclusion: The decreased ∆R1 slopes of O2 enhancement in arterial blood and lung after smoking detected by our O2-enhanced T1 mapping reflect a decreased O2 transfer (2). This could be potentially explained by an increased perfusion as observed in our spin labeling experiments.

Perfusion References: 1. Wang T, et al. ISMRM, Toronto, 2003. 2. Jakob PM, et al. MRM, submitted. 3. Wang T, et al. JMRI, in print.

346 Characterisation of systematic error in SVD based deconvolution for quantitative perfusion measurements T. K. Carpenter, J. M. Wardlaw; School of Molecular and Clinical Medicine, Clinical Neurosciences, Western General Hospital, Edinburgh, UNITED KINGDOM. Introduction: Typical methods for measurement of absolute cerebral blood flow (CBF) and mean transit time (MTT) using dynamic susceptibility contrast MRI require the deconvolution of an Arterial Input Function (AIF) and the acquired tissue concentration-time curves. One technique uses singular value decomposition (SVD) to perform the deconvolution [1] and this work characterises the systematic error resulting from physiological CBF and MTT values as found in stroke. Methods: A signal model described in detail elsewhere [1] was used to produce simulated concentration-time curves based on those observed using our imaging hardware and protocol. The model parameters were varied over the range CBF 0.0171ml/100ml/s, CBV 0.33 to 5 % and MTT 2 to 20s. 256 MTT-CBF combinations were chosen (true pairs) to give a uniform sampling of the possible MTT-CBF values dictated by the range of CBV values. An exponential residue function was assumed, no delay between the AIF and tissue concentration-time curves was modelled and additive noise was not used, so that only systematic error in the parameter estimates (estimated pairs) remained. The deconvolution was performed under two different conditions: an optimum global threshold value (as in [1]) and an optimum local threshold found for each known MTT-CBF pair. The optimum threshold was defined as minimising the magnitude error in the MTT-CBF pairs. Results: Figure 1 shows the estimates obtained using a global threshold to be compressed towards the centre of the CBF-MTT scatter plot, whilst those obtained with a local threshold (Figure 2) are in better agreement with the true values. However, the error is still non zero for longer MTT. Figures 3 and 4 show the relationship between the optimum local threshold and MTT and CBF values, respectively. Discussion: In patients with cerebral infarction long MTT values would be expected in the lesion. The results of this study suggest that within this region the values measured using a global threshold may be unreliable. Due to intrinsic limitations of the SVD deconvolution, the threshold that reduces the error to zero cannot always be obtained particularly for longer MTT values. Figures 3 and 4

S183

suggest that the local threshold is a function of MTT and independent of CBF indicating that choosing an acceptable threshold for a particular voxel requires prior knowledge of the MTT. References: 1. Ostergaard L et al [1996], MRM; 36:715-725

S184

Perfusion single-shot multi-echo EPI acquisition and phase contrast MRI during CO2 inhalation in patients is possible. The discrepancy between vascular territories perfused by an impaired MCA and less extensive area of reduced BOLD response may indicate collateral blood supply. These differences reflect the different underlying parameters: In TCD or PC only flow in major arteries is detected directly, but no measurement of tissue perfusion is obtained. It is still to be examined which parameter has the highest predictive value for clinical prognosis.

347 Quantitative measurements of the perfusion reserve using transcranial Doppler ultrasound and phase contrast and BOLD MRI O. Speck1, S. Ziyeh2, J. Rick1, A. Hetzel3, J. Hennig1; 1Dept. of Diagnostic Radiology, Section Medical Physics, Freiburg, GERMANY, 2Dept. of Neurosurgery, Section Neuroradiology, Freiburg, GERMANY, 3Dept. of Neurophysiology, Doppler Laboratory, Freiburg, GERMANY. Introduction: In patients with neuro-vascular disease, the perfusion reserve capacity is an important predictor for the risk for ischemic events. Transcranial Doppler sonography (TCD) during inhalation of a CO2-air gas mixture to determine flow changes in the MCAs is the method most commonly employed clinically. Recently, BOLD methods have been proposed to assess the tissue perfusion reserve capacity. However, discrepancies between the results of both methods have been reported. We have added quantitative phase contrast flow measurements (PC) during CO2-inhalation to the MR protocol in order to directly compare the flow increase in the feeding arteries (MCA) and the tissue BOLD effect. The goal was to resolve possible discrepancies and exclude possible non-reproducibilities between the TCD and MRI examinations. Subjects and Methods: Patients with different degrees of unilateral stenosis of the common or internal carotid artery and normal controls were included. For TCD measurements, both middle cerebral arteries (MCA) were insonated through the temporal bone window. Flow was measured before and during the inhalation of 7% CO2-air mixture. The MR-protocol started with T1, T2 imaging and tof-MRA. ECG gated quantitative PC measurements before and during CO2 inhalation in 2 slices perpendicular to the MCAs and quantitative BOLD MRI (single-shot multi-echo EPI) were performed on a 1.5T Siemens Sonata. For the BOLD measurement of 5 minutes duration, the patients were breathing the 7% CO2-air mixture during 2 minutes. Results: An example of CO2 induced flow increase in the MCAs of a patient is shown in figure 1. Quantitative PC and BOLD MRI was successfully applied to and well tolerated by all patients and controls. All controls showed symmetrically increased BOLD (gray matter mean 34%) and PC (MCA mean 41%) perfusion during CO2. In some patients the results from PC and BOLD measurements differed, but PC and TCD results agreed well. Discussion: Quantitative mapping of the perfusion reserve with

348 Comparison of cerebral blood volume maps generated from T2*- and T1-weighted first pass dynamic contrast-enhanced MRI data H. A. Haroon1, T. A. Patankar1, X. P. Zhu2, K. L. Li2, A. Jackson1; 1Imaging Science and Biomedical Engineering, University of Manchester, Manchester, UNITED KINGDOM, 2Department of Radiology, University of California at San Francisco, San Francisco, CA. Purpose: The close relationship between histological grade and MR-based cerebral blood volume (CBV) maps in gliomas make them desirable for surgery planning and guidance. CBV maps are usually obtained from T2*-weighted data (T2*-CBV) [1], acquired using EPI sequences. We have calculated CBV maps using a First Pass model [2] and contrast-enhanced dynamic T1-weighted data (T1-CBV) acquired over a short time. Our model calculates estimates of the volume transfer constant Ktrans and CBV from the same data using a decomposition process, removing the effects of extravascular contrast leakage on the CBV estimates. We have compared our T1-CBV maps with T2*-CBV maps in terms of their spatial resolution and median measurements from tumour. Subjects and Methods: 9 patients with confirmed cerebral neoplasms (8 gliomas, 1 metastasis) were recruited for this study. Imaging was performed at 1.5T. T1-CBV maps were calculated using 3D-T1W-FFE sequence (128×128×25 matrix, 230mm fieldof-view, 3mm slice thickness, TR=4.2ms, TE=1.2ms) with 3 precontrast data sets acquired at flip angles of 2º, 10º and 35º; followed by a dynamic contrast-enhanced (0.1mmol/kg of GdDTPA-BMA) acquisition at 35º (20 scans every 5s) [2]. T2*-CBV maps were calculated using 4-shot 2D-T2*W-FEEPI sequence

Perfusion (128×128×9 matrix, 230mm field-of-view, 6mm slice thickness, TR=440ms, TE=30ms) and contrast-enhanced (second single dose) acquisition at 35º (52 scans every 1.8s) [3]. Regions-of-interest were drawn manually on visually matching slice positions and orientations (such as Fig.1 and Fig.2). Results: Median T1 and T2* CBV in enhancing tumour ROIs (Fig.3) showed good correlation (R=0.65,P<0.05). A visually appreciable spatial correlation between the maps is also seen. Blood vessels appear larger than their actual size on T2*-CBV maps when compared with high resolution anatomical images, due to paramagnetic effects on perivascular T2*. T1-CBV maps show poor signalto-noise in grey and white matter, but this improves in higher CBV areas such as tumour.

Discussion/Conclusion: We show that our leakage-free T1-CBV maps correlate well with T2*-CBV maps, both in terms of median tumour CBV and visual distribution of blood vessels. This supports our view that the same biological information can be gained as from T2*-CBV maps. T2*-CBV maps, however, are not corrected for extravascular contrast leakage, possibly causing the scatter in Fig.3. T1-CBV maps do not suffer from the artefacts seen in T2*CBV maps, offer higher spatial resolution, and are thus preferable for surgical planning and guidance. References: 1. Maeda M, et al.(1993)Radiology 189:233-238 2. Li KL, et al.(2000)J.Magn.Reson.Imaging 12:347-357 3. Zhu XP, et al.(2000)J.Magn.Reson.Imaging 11:575-585

S185

349 A modified perfusion analysis for the evaluation of tissue viability following cryotreatment of mice tumors C. T. Brami, PhD, D. Manor, MD, PhD, Y. Itzchak, MD, PhD, T. Kushnir, PhD; Department of Diagnostic Imaging, The Chaim Sheba Medical Center, Tel Hashomer, ISRAEL. Purpose: We propose to develop a method to evaluate tumor destruction by quantifying the hemodynamic status following cryotreatment. The method is based on a novel first-pass analysis and may serve as an early in vivo NMR marker for the success of treatment. Methods: Animals: Female mice bearing DA3 tumor were used for the study. Cryotherapy: Tumor was frozen using a 2.5 mm cryoprobe (Galil Medical Ltd), which was inserted into it percutaneously. Ablation protocol: freezing at -120oC for 1 minute, waiting 5 min for slow thawing, freezing again 1minute for a second cycle. A one centimer iceball was created in the tumor. MR Imaging: MRI data were acquired using a 3 Tesla magnet (Signa Horizon LX, GE) and a small volume coil. 1) First-pass study: A single 1 mm slice was acquired using a fast GRE sequence for BOLD and flow contrast: Time resolution 0.6 second, TR/TE/flip angle 16.7/11/45o, resolution 0.6x0.3 mm2. The contrast agent Gd-DOTA was injected into the mouse vein tail after 1minute of the 2minutes dynamic study. 2) Gadolinium-enhancement: T1weighted 3D SPGR images, acquired before and after the first-pass study, were subtracted: TR/TE/flip angle 15.5/3.3/20 o, resolution 0.3x0.3 mm2. Results: In the first-pass study some pixels demonstrated biphasic rapid signal intensity changes following Gd injection (fig 1): first a deep due to Gd pass into the vessels (T2* contrast), followed by T1-dependent increase due to Gd leakage from vessels. The timecourse was fitted by a sum of two gamma functions: one having a negative peak and another, time-shifted relative to the first, having a positive peak. Fit was achieved using standard minimization routines implemented in Matlab. Pixel by pixel values of the sum of squares of the differences between data and the function, divided by the squared standard deviation of the data and inversed, (inversed normalized least square error) produced a map reflecting the quality of fit. The resulting map demonstrated well-defined regions localized only around the tumors. This combination of T2* (representing perfusion) and T1 (representing leakage) may lead to better definition of viable tumor tissue. Although the conventional postcontrat substraction image represented better the vascular stasis in the ablated region ; in our model, the viable regions of the residual tumor were better enhanced. Conclusion: A new analysis of dynamic, mixed T2* - T1 contrast images, is proposed that could display specifically perfusion related to viable tumor.

S186

Brain, Ischemia

POSTER PRESENTATIONS Brain, Ischemia 350 The use of fast FLAIR for the early diagnosis of the brain infarction at 0,15 T MR - Tomography L. V. Gubskii, N. A. Shamalov, K. V. Sokolov, N. V. Mazov, A. N. Shibaeva; Chair of Neurology and Neuroradiology, Russian State Medical University, Moscow, RUSSIAN FEDERATION. Purpose: The aim of this study was to assess the contrast of ischemic brain lesions (IBL) by FLAIR within the 72 hours after stroke onset. Subjects and Methods: 103 patients with hemispheric infarction (aged from 48 to 89 years old, 50 males and 53 females) have been studied, 82 of them have been studied twice during this period. The study was performed at the 0.15 Tesla MR-scanner (Ellipse, Az, Russia). Used pulse sequences were: fast FLAIR (TR = 4100, TE = 100, TI = 1155), TSE (T2-weighted), SE and GRE (T1- weighted). Contrast (C) of IBL was calculated in relation to the intact white matter (wm) as: C = (SIibl - SIwm) / SIwm, where SI - signal intensity. Results: In 6 (5,8%) patients infarction was not visible during the first day after stroke, 3 of them had lacunar infarction and have been studied at 6, 21 and 22 hours and other 3 patients have been studied at 4, 8 and 10 hours after stroke onset. At later studies IBL were found out in all of these patients. In 97 patients (94,2%) the values of contrast for FLAIR were positive and gradually increased from 4 to 24 hours after stroke (Fig. 1).

Fig. 1. Contrast of IBL on FLAIR. * - p<0,05, ** - p<0,01 in comparison with the study at the 4-8-th hours, # - p<0,01 in comparison with the study at the 24-36-th hours. All IBL within 12 hours were found in gray matter (cortex and basal ganglia) that can be reflection of the cytotoxic edema. In 24 36 hours the contrast of IBL increased significantly in comparison with the study at the 4 - 8-th hours and in 48-60 hours these differences were more significant. On T2-weited images the conspicuity of the IBL were worse than on FLAIR because the signal from cerebrospinal fluid didn’t allow assessing the IBL in some cases. On T1-weited images the values of contrast of IBL were low negative (without significant differences). Conclusion: On 0,15 T scanner FLAIR is the more sensitive method for early diagnosis and assessment of brain infarction than conventional T1- and T2-weited images.

351 Diffusion-weighted imaging on low-tesla MR unit in early stroke – the method and diagnostic value R. Krawczyk1, R. Poniatowska1, J. Ryterski1, R. Boguslawska1, A. Bochynska2, A. Kobayashi2, M. Chahwan2, Z. Lysiak2; 1Magnetic Resonance, Institute of Psychiatry and Neurology, Warsaw, POLAND, 2Neurology, Institute of Psychiatry and Neurology, Warsaw, POLAND. Purpose: The aim of the study was to assess the usefulness of diffusion-weighted imaging in early stroke diagnose on low-tesla magnetic resonance unit. Material and Method: 40 patients (24 men and 16 women, mean age 73,1+/- 11,1 years) with clinical symptoms of early stroke and normal CT image were studied using 0,23T magnetic resonance unit. The time between the onset of the symptoms and MR study varied from 1 to 22 hours. The MR study consisted of axial FLAIR and Diffusion-weighted scans. The slice thickness on FLAIR scan was 5 mm, on DWI 10mm. The total time of the study was less than 10 minutes. The DWI scan was performed 2 times – with b-value 0 and 600 to perform Apparent Coefficient Map (ADC map). On ADC map the standardized diffusion coefficient was measured, as the ratio between the value of the elliptic ROI of the lesion and of the collateral brain tissue. The volume of the lesion was manually calculated as the sum of the areas of the foci on all slices multiplied by slice thickness. Results: The total number of lesions found on MR studies was 49. The volume of the foci varied from 0,3 to 233,2 cm³ (21,26 cm³ +/42,85 cm³). The patients with the biggest lesion volume had very poor clinical outcome. In 16 cases the lesion was not yet seen on FLAIR images indicating very early stroke. In 25 patients FLAIR sequence showed multiple ischemic foci, in 2 encephalopathy, in 8 single ischemic lesions and in 5 no lesions beside infarct were found. In 13 cases the new lesion was seen in the area of previous stroke. In 2 cases the new stroke coexisted with old infarct lesion located in different part of the brain. The standardized diffusion coefficient in new stroke area varied from 0,867 to 0,987(0,934+/0,029), in old stroke was elevated – from 1,056 to 1,148(1,087+/-0,034). In 2 cases was not possible to calculate ADC map due to patients movement between DWI scans. Conclusion: The diffusion-weighted imaging on low-tesla unit is very useful diagnostic tool in early stroke stage. This method allows to assess not only the localization, but also the size of the lesion, which can be valuable clinical predictive factor.

352 Wallenberg's lateral medullary syndrome: Clinical and Diffusion-weighted imaging findings O. Kitis1, C. Calli1, H. Altay1, N. Yunten1, A. Kocaman2, H. Sirin2; 1Radiology, Ege University Medical School, Izmir, TURKEY, 2Neurology, Ege University Medical School, Izmir, TURKEY. Purpose: The aim of the study is to review the most common clinical findings and to investigate the efficacy of diffusion-weighted imaging in the patients with Wallenberg’s lateral medullary syndrome. Subjects and Methods: 13 patients with Wallenberg’s lateral medullary syndrome were examined with conventional MR images and echoplanar diffusion-weighted in 1.5 T MR unit. MR examinations were obtained in acute or subacute stage of clinical syndrome

Brain, Tumours

S187

and diffusion-weighted imaging was considered to be positive for infarction when an increase in signal was seen on b=1000 s/mm2 images in the posterolateral medullary localization. Results: The most common neurological symptoms initially encountered in our patients were ataxia, vertigo, numbness, disartria, nausea, vomitting and disphagia. Neurological sings were as ataxia, horner's syndrome, contralateral hypalgesia of the body, ipsilateral facial hypalgesia and nystagmus. Diffusion-weighted images were positive in 12 patients in the acute or subacute stages of this clinical syndrome, false-negative DWI obtained in only 1 patient who was examined within the first day, 10. hour after the onset of symptoms. In the visual evaluation of the diffusion-weighted imaging, the contrast between normal and infarcted brain stem area was better in the high b value images than the ADC (apparent diffusion coefficient) map images. Conclusion: Diffusion weighted imaging is a valuable technique in the examination of patients presenting the signs and symptoms of Wallenberg's syndrome and high b value images can provide complementary data to T2-weighted images.

discriminate all regions at each time points. Significant differences were found except for rTTP at day 4 and day 120 (p=0.7 and 0.5). Conclusion: The knowledge of the evolution of imaging parameters can be helpful in the understanding of stroke mechanisms. It can also be of great interest in the use of a thrombolitic therapy.

353

Brain, Tumours

Time Evolution of PW and DW MRI parameters in acute and chronic stroke F. C. G. Schneider, Jr.1, A. Januel1, L. Court1, P. Garnier2, M. Homel3, A. Jaillard3, S. Laporte4, D. Michel2, F. Barral1; 1Radiology, CHU Saint Etienne, Saint Etienne, FRANCE, 2Neurology, CHU Saint Etienne, Saint Etienne, FRANCE, 3Neurology, CHU de Grenoble, Grenoble, FRANCE, 4Médecine thérapeutique, CHU Saint Etienne, Saint Etienne, FRANCE. Purpose: We investigated the time evolution of the parameters measured by diffusion weighted (DWI) and perfusion weighted (PWI) MRI from the acute to the chronic stage. Subjects and Methods: 101 patients were included in this study within 24h of symptom onset (mean/std/median 12/6/12h). 95 underwent a T2 follow up examination at day 4 and 77 120 days later. 69 patients with a visible stroke on follow up are investigated here. The lesion volumes were manually delineated by 2 experimented neuro-radiologists on the initial DW, PW (TTP) and follow up images. Images were co-registered altogether using the normalised mutual information algorithm of SPM2b (FIL, London). Four regions were derived from the delineations : i) the core of the infarct was defined as the intersection of initial DWI and follow up image, ii) a positive infarct growth (IG+) area defined as present on follow up image whereas not on initial DWI, iii) a negative infart growth (IG-) area defined as present on initial DWI and not on follow up, iiii) an oligaemic region defined as present on initial PWI and not on follow up not initial DWI. Relative apparent diffusion coefficient (rADC), relative blood volume (rBV) and relative time to peak (rTTP) were recorded in the delimited areas for each patients / groups. ANOVA and t-tests were used to compare these areas at each time point. Results: The average measured volumes of the core, IG-, IG+ and oligaemia were 35.7, 12.9, 31 and 35.2 cm3 receptively. Initially, the core was defined by reduced rADC, rBV (different from 1, p<0,0001) and increased rTTP (different from 0, p<0,0001). No significant differences were found between day 1 and 4 for the core rADC and between day 4 and day 120 rTTP. The rADC of IG+ was not different from 1 for day 1 (p=0.07) as well as rBV for day 1 and day 4 (p=0.3 and 0.3 respectively). Oligaemia had a rADC close to 1 (0.91) but significantly different (p<0.005) and a rBV increased for day 1 and 4 (p<0.05 and <0.01). ANOVA models were tested to

354 Application of spectral deconvolution to in vivo MR spectroscopy: Evaluation of brain tumors H. Kim1, S. Eun2, H. Kang, MD1, J. Seo, MD1, G. Jeong, PhD1; 1Department of Radiology, Chonnam National University Medical School, Kwang-Ju, REPUBLIC OF KOREA, 2Department of Medical Engineering, Chonnam National University, Kwang-Ju, REPUBLIC OF KOREA. Introduction: Direct application of in vivo 1-H MRS to a biological system is restricted by a lack of spectral resolution due to line broadening and extensive peak overlaps accompanied by low SNR. In this study a spectral deconvolution technique was used to increase accurate interpretation of cerebral metabolic changes from 1-H MR spectrum in grading of tumor malignancy as well as differential diagnosis of brain tumors. Subject and Methods: Twenty-two patients pathologically confirmed were examined with a 1.5T Sigma MR scanner for assessment of astrocytoma and meningioma. A STEAM pulse sequence (TR/TE/MT= 3,000/30/13.7msec) with 96 scans was used. Postprocessing of the MRS was performed by MRUI program (http://www.mrui.uab.es). The creatine (Cr) peak was used as a standard to normalize signal intensities in all 1-H MR spectra. Results: Combined use of in vivo 1-H MRS and spectral deconvolution increased overall quantitative accuracy of the metabolite concentration by 70%. In the ow-grade astrocytoma, concentration of NAA was reduced by 24% and Lac was markedly elevated by as much as 437% below the normal values; in the high-grade astrocytoma, NAA was reduced by 31%, and Lac, Cho, mI, and alpha-Glx were elevated by 592%, 142%, 131%, and 99%, respectively(P<0.05). In a differential diagnosis between low- and high-grade astrocytoma, elevation of both Cho and mI is significantly related with high-grade(P<0.05); while elevation of alphaGlx, low-grade astrocytoma. The pattern of metabolite changes in meningioma was similar to that of high-grade astrocytoma in that NAA was reduced by 38% and Lac and Cho were elevated by 470% and 120%(P<0.05), respectively. However, the pattern of mI, alpha-Glx, and beta and gamma-Glx were different from each other; especially the levels of mI and alpha-Glx were

S188

Brain, Tumours

significaly(P<0.05) important for differential diagnosis between them. In a similar way with high-grade astrocytoma, the cyst component associated with high-grade astrocytoma showed the same pattern of concentration changes: reduction of NAA by 21% and elevation of Lac, alpha-Glx, mI, and Cho by 236%, 100%, 88%, and 16%, respectively, at the significance level with P<0.15. However, overall levels of metabolite changes of cyst are smaller than those of high-grade astrocytoma. Conclusions: Combination of in vivo 1-H MRS and a spectral deconvolution technique led to more accurate diagnosis and grading of meningioma and astrocytoma, as compared with direct application of 1-H MRS only.

References: 1. Barba I., Cabañas M.E., Arús C., Cancer Research 59, 18611868, 1999 2. Moreno A., Arús C., NMR in biomedicine 8, 33-45, 1996

355 Myo-inositol and glutamine display growth curve dependent changes in rat glioma C6 cells D. Valverde1, A. Candiota1, M. Cabañas2, M. Quintero1, L. Badiella3, C. Arús1; 1Departament de Bioquímica, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, SPAIN, 2Servei de RMN, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, SPAIN, 3Servei d'Estadística, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, SPAIN. Purpose: The objective of this study was to analyze in C6 rat glioma cells, as a model of an aggressive glial tumour, if there were quantitative changes in its 1H NMR spectral pattern that could correlate with proliferation rate in the different phases of their growth curve in cell culture. Experimental: Perchloric acid (PCA) extracts of about 108 C6 ATTC cells at log, confluence and postconfluence (cells were cultured essentially as described in [1]) were performed in order to extract water soluble metabolites (n = 3 for each growth phase)[2]. Protein content in the PCA insoluble pellet was analyzed with the Lowry method. PCA soluble metabolites were resuspended in 400µ l of PBS-D2O and spectra recorded at 9.4 T (Bruker ARX, presat pulse sequence, 90º pulse, 30 sec TR). TSP was used as a chemical shift and quantitation reference while exogenous fumarate was used to correct for possible extraction losses. Spectra were analyzed by deconvolution using WinNMR 6.1.0.0 (Bruker Daltonik, Gmbh). Statistical analysis of the metabolite concentration was performed using ONE-WAY ANOVA test and with a General Lineal Model (GLM) using SPSS 10.0.6 (SPSS Inc.) software package. Statistical analysis of the protein concentration was carried out with a non parametric Kruskal-Wallis test using SPSS. Statistical level of significance was set at p<0.05. Results: Metabolites quantified were glycine, taurine, GABA, creatine, lactate, glutamate, alanine, phosphocholine, glycerophosphocholine, myo-inositol and glutamine. Protein was also quantified. Significant changes (table 1) were only detected for myo-inositol and glutamine. Table 1 and figure 1 show major quantitative changes detected Discussion: Most metabolites did not show statistically significant differences due to the proliferation state of the cells. Nonetheless, myo-inositol and glutamine did. Myo-inositol concentration almost doubled when proliferation rate slowed down between log phase and post-confluence. On the other hand, glutamine concentration almost halved. Both resonances are detectable in vivo in human tumours at short echo times (TE 20-30 msec). Then, detection of changes in the in vitro pattern for these two metabolites may have potential as proliferation marker, provided C6 results here described can be extrapolated to human tumours in vivo.

Major quantitative changes detected Phase

Metabolite (µ mol/g

Protein

wet weight) ± SD#

(mg/g wet weight)&

Myio-inositol Glycine

Creatine

Log (n=3), 3 days

3,94± 0,76

3,72± 0,69$ 1,29± 0,25 10,96± 2,01

Confluent (n=3), 5 days

3,59± 0,55

2,22± 0,66 0,92± 0,71 19,63± 6,36

Post-confluent (n=3), 7 days 7,78± 1,27*

2,05± 0,47 1,11± 0,24

27,55± 1,72

Label for statistically significant differences between (*) post-confluent and other ($) log and other. #Concentration is given per g of wet weight of a packed pellet of C6 cells. & Because SD was different for protein content between different growth phases according to a Levene test, a non-parametric Kruskal-Wallis test had to be carried out instead of a ONE WAY ANOVA, and no statistically significant differences were found.

356 Contribution of sialic acid in macromolecules to the ex vivo 1H MRS pattern of cystic fluid in human brain tumors A. P. Candiota1, M. R. Quintero1, M. Cabañas2, C. Aguilera3, J. J. Acebes4, C. Arús1; 1Bioquímica i Biologia Molecular, Universitat Autònoma de Barcelona, Cerdanyola del Valles, SPAIN, 2SeRMN, Universitat Autònoma de Barcelona, Cerdanyola del Valles, SPAIN, 3Institut de Diagnòstic per la Imatge (IDI), Hospital Duran i Renyals, CSU de Bellvitge, Hospitallet de Llobregat, SPAIN, 4Department of Neurosurgery, Hospital Prínceps d'Espanya, CSU de Bellvitge, Hospitalet de Llobregat, SPAIN. Introduction: The resonance at about 2.03 ppm in in vivo spectra of brain tumour cystic lesions has been attributed to NAA from

Brain, Tumours outer volume contamination[1]. Nevertheless, this resonance is also observed in ex vivo spectra of cystic fluids, and is poorly recovered in PCA extracts. We have investigated seven fluid samples to better understand the origin of the in vivo pattern. Subjects and Methods: Samples were collected from patients undergoing surgery. Informed consent was obtained in all cases. Acquisition parameters for in vivo (1.5T) spectra were TE 136ms, TR 2000ms, 512 points and SW 1000Hz. For ex vivo spectra (9.4T), a "jump and return" water supression sequence was used with TE 136ms, TR 10s, 16K points and SW 4807Hz. For in vitro PCA extract spectra TR was 30s. The signal of interest was quantified using TSP as reference. PCA extract of the fluid was carried out as in [2]. Sialic acid (SA) and hexuronic acid-containing compounds were quantified as in [3,4]. Results: We have analysed 7 fluid samples and the 2.03 ppm resonance was detected in all of them (fig.1). A concentration of 4.66±1.21mM was calculated assuming that the signal originates in CH3 groups and was only partially extracted by PCA (mean recovery: 28.3±9.7%). Ex vivo quantitation was made in TE=136ms spectra, taking into account corrections for the measured T2 (biexponencial) as T2,a=48.4±11.4ms and T2,b=649.6±101.3ms. SA chemically analyzed was 2.09±0.53mM, which is about 45% of the amount detected by ex vivo 1H-MRS. Hexuronic acid-containing compounds accounted for 15% of the resonance. Discussion: Our results are compatible with a PCA-insoluble macromolecule with N-acetyl groups being responsible for most of the 2.03 ppm in vivo resonance. The main contributor seems to be a macromolecule-linked SA. Other authors have described a similar in vivo signal incompatible with an NAA origin [5]. Besides, it is known that tumour cells shed sialyl-containing compounds to the extracellular space [6]. About 15% of the signal may be contributed by chondroitin-sulphate type glucosaminoglycans, another type of molecules shed by tumor cells. 1. Chang K-H et al.[1998]. AJNR 19:401-405 2. Remy, C. et al.[1994] J. Neurochem. 62: 166-179 3. Jourdian G.W et al. [1971] J Biol Chem 246: 430-435. 4. Bitter, T. et al. [1962]Anal Biochem. 4: 330-334 5. Burtscher, I. et al.[2001]. J Magn Reson Imaging 13:560-567 6. Heaney-Kieras, J. et.al.[1986] J Natl Cancer Inst 77:643-648

S189

357 Usefulness of single-voxel proton MR spectroscopy and 1HMR spectroscopic imaging (MRSI) for diagnosing of brain glial tumor recurrence E. Skrobowska1, A. Warczynska1, S. Skrzynski2; 1Department of Radiology, Medical Military Institute, Warsaw, POLAND, 2Department of Neurosurgery, Medical Military Institute, Warsaw, POLAND. Introduction: Conventional magnetic resonance imaging (MRI) is commonly employed to assess recurrence of brain tumors. In some cases, especially when there is hyperintensity on FLAIR and no contrast enhancement around the resection cavity, there are great difficulties in interpretation. Purpose: The aim of this study was to evaluate usefulness of proton MR spectroscopy in diagnosis of tumor recurrence in patients after neurosurgical treatment. Subjects and Methods: The present study included 52 healthy volunteers and 60 patients with gliomas (24 with primary gliomas and 34 with suspected recurrence after radical tumor resection). In all 60 patients MRS diagnosis was confirmed histologically. The studies were performed on a 1.5 T GE Signa LX scanner using a commercially available package PROBE 2000. After the routine MRI scan (before and after Gd-DTPA), single-voxel spectra were acquired using PRESS, 35TE and 1H-MR spectroscopic imaging using PRESS, 144TE. We calculated referral values of metabolites ratios (NAA/Cr, Ch/Cr, mI/Cr, Ch/NAA, NAA/Ch) in groups of healthy volunteers, in patients with primary gliomas and with recurrent tumors. The Mann-Whitney U-test was used to verify statistical significance between the metabolites ratios obtained for three groups of patients. Results: The mean values of metabolites ratios in groups are shown in table1. Comparison between groups (Mann Whitney Utest) are shown in table 2. Table 1. Group

No NAA/Cr of patients

Ch/Cr

mI/Cr

Ch/NAA

NAA/Ch

Healthy

52

1.50 ±0.16 0.83 ±0.09 0.62 ±0.18 0.56 ±0.07 1.83 ±0.27

Primary glioma

24

1.05 ±0.31 1.70 ±0.51 0.95 ±0.25 1.87 ±1.25 0.69 ±0.3

Recurrent glioma

36

1.09 ±0.26 1.78 ±0.65 0.99 ±0.22 1.70 ±0.62 0.68 ±0.31

Table 2. p value healthy/ primary glioma

NAA/Cr 0,000000

Ch/Cr 0,000000

mI/Cr 0,000000

Ch/NAA 0,000000

NAA/Ch 0,000000

healthy/ recurrent glioma

0,000000

0,000000

0,000000

0,000000

0,000000

primary glioma/ recurrent glioma

0,785926

0,506741

0,629200

0,586986

0,586986

The high choline signal measured by 1H-MR spectroscopic imaging was observed in areas of tumor recurrence, despite the absence of contrast enhancement.In the group of 36 patients with suspected tumor recurrence, MRS diagnosis was confirmed histologically in all cases. Conclusions: Single-voxel proton MR spectroscopy and 1H-MR spectroscopic imaging (MRSI) represent complementary techniques and are able to diagnose tumor recurrence. MRS have to be routine method in the follow-up of patients with neurosurgically treated gliomas.

Brain, Tumours

S190

358 A study of 66 patients with cerebral glioma to determine whether it was possible to distinguish between oligodendroglioma and oligoastrocytoma using metabolic ratios from single voxel proton MR spectroscopy E. T. S. Smith1, K. Tyler1, K. Das1, P. Cole2, J. Broome1; 1Neuro X-Ray Department, The Walton Centre NHS Trust, Liverpool, UNITED KINGDOM, 2Department of Physics, The University of Liverpool, Liverpool, UNITED KINGDOM. Introduction: For some years now, at our Institution, single voxel proton MR Spectroscopy of the brain has been incorporated along with 2D MRSI into the routine work up of patients with suspected cerebral glioma. The aim is to confirm, if possible, the presumed diagnosis of glioma and also to identify areas of increased metabolic activity prior to a biopsy procedure. A further advantage would be if MRS could suggest likely tumour grading and if it were to be possible to distinguish oligoastrocytoma from oligodendroglioma, since the latter may respond better than the former to certain chemotherapeutic agents. Purpose: The purpose of the study was to ascertain if examination of commonly obtained metabolite ratios would assist in grading the gliomas and also to ascertain if it was possible to distinguish between oligoastrocytoma and oligodendroglioma on the basis of metabolic profiles as detected by SV proton MRS. Subjects and Methods: Over a 3 – 4 year period, 66 patients with histologically proved cerebral glioma have been studied. At the time of the initial MRS examination, no treatment had been given. The MRS and MRI examination preceded or followed stereotactic biopsy within six weeks. MRI and MRS were performed on a 1.5T GE Signa Horizon LX Nvi MR Scanner. Spectra were obtained using a single voxel technique, 2cm x 2cm x 2cm in size. A PRESS sequence was used, TR 1500ms TE 35ms. Acquisition time was 2 minutes and 6 seconds or 3 minutes and 46 seconds depending on the software in use at the time. The study usually followed Gadolinium enhancement which was used as a guide to voxel placement – the voxel being located over the site of maximum enhancement. To obtain an internal standard, a second spectrum was obtained from a voxel placed over the contralateral area of the brain. The manufacturer's software only was used for analysis – the aim being to assess the information value of the spectra gained in a clinical setting without additional post processing software. Results: Ratios for normal white matter mI/Cr

Cho/Cr

0.64 +/- 0.3

1.03 +/- 0.41 1.46 +/- 0.66 2153 +/- 570

NAA/Cr

H20/Cr

Ratios for oligoastrocytoma Grade

mI/Cr

1(n=6)

0.95 +/- 0.22 1.23 +/- 0.1

Cho/Cr

NAA/Cr

H20/Cr

1.33 +/- 0.2

2 ( n = 30 )

0.85 +/- 0.3

4690

3(n=6)

0.41 +/- 0.08 1.48 +/- 0.44 0.91 +/- 0.31 5774 +/- 1822

4(n=2)

0.41 +/- 0.09 1.29 +/- 0.31 0.84 +/- 0.05 7600

1.45 +/- 0.56 1.02 +/- 0.58 3485 +/- 1448

Ratios for oligodendroglioma Grade

mI/Cr

Cho/Cr

2 ( n= 18 )

0.82 +/- 0.3

1.36 +/- 0.36 0.99 +/- 0.48 3726 +/- 912

NAA/Cr

3 ( n= 4 )

0.61 +/- 0.16 1.63 +/- 0.03 0.96 +/- 0.16 3718 +/- 234

359 The value of diffusion-weighted imaging in intracerebral tumors C. Calli1, O. Kitis1, H. Altay1, T. Yurtseven2, N. Yunten1; 1Radiology, Ege University Medical School, Izmir, TURKEY, 2Neurosurgery, Ege University Medical School, Izmir, TURKEY. Purpose: To investigate the role of diffusion-weighted imaging (DWI) in the diagnosis and differential diagnosis of intracerebral tumors. Subjects and Methods: 48 patients, ages ranging from 14 to 65 years, with malignant intracerebral tumors were included in the study. All the patients were examined by MRI using turbo spinecho T1W, T2W and contrast enhanced sequences. Besides the standard MR imaging, DWI using echo-planar imaging (TR/TE:4700/18msec) was performed in all the patients with b values of 0, 500, and 1000sec/mm2 in three orthogonal axis. ADC (apparent diffusion coefficient) map images were generated automatically by the MR Unit. ADC values were calculated. The diagnosis were proven either by surgery or stereotactic biopsy. Results: According to the pathological findings there were 21 high grade gliomas (grade 3-4) (glioblastoma multiforme=8, anaplastic astrocytoma=9, anaplastic oligodendroglioma=4), 9 low grade gliomas (grade 2), 7 lymphoma and 11 metastasis in the study group. In 19 of the patients contrast enhancement was not present in the tumor (grade2=9, grade 3=10). The ADC values of the tumors were calculated as follows: high grade gliomas=0.92 ± 0.28 x 10-3 mm2/s, low grade gliomas=1.42 ± 0.26 x 10-3 mm2/s, lymphomas=0.85 ± 0.18 x 10-3 mm2/s, and metastasis=0.90 ± 0.21 x 10-3 mm2/s. The ADC values (calculated from the solid portion of the tumor) of high grade gliomas and metastasis did not differ from each other. Lymphomas showed slightly lower ADC values than other tumors. However, ADC values enabled the differentiation between non-enhancing high and low grade gliomas. Conclusion: DWI with the calculation of the ADC values provides additional information in the differential diagnosis of intracerebral malignant tumors.

H20/Cr

Discussion/Conclusion: In oligoastrocytoma, there appears to be a trend towards a lower mI/Cr ratio with higher grade of malignancy. The trend is less marked with oligodendroglioma but numbers are small. There appears to be no significant difference between the measured metabolite ratios of oligoastrocytoma and oligodendroglioma in this series.

360 MRI usefulness in the assessment of neurosurgical treatment efficacy and pituitary function in patients after transsphenoidal pituitary adenomectomy A. Warczynska1, E. Skrobowska1, G. Zielinski2, A. Koziarski2; 1Department of Radiology, Medical Military Institute, Warsaw, POLAND, 2Department of Neurosurgery, Medical Military Institute, Warsaw, POLAND.

Brain MRS Purpose: The aim of the study was to evaluate the MRI usefulness in the assessment of the neurosurgical treatment effects in patients after transsphenoidal pituitary adenomectomy and to estimate the correlation between image and function of hypophysis. Material: 70 patients (46 female and 24 male), aged 16 – 76 years, operated for pituitary adenomas with the transsphenoidal approach. There were 34 cases of acromegaly, 9 prolactinoma, 12 cases of Cushing disease, 4 thyreotropinoma, 3 gonadotropinomas and 8 non secreting adenomas. Methods: In all the patients before and after transsphenoidal adenomectomy the hypophyseal hormonal function and field of vision were compared with the MRI assess of the anterior and posterior pituitary lobes, stalk and optic chiasm. MRI of the parasellar region was performed with a 1,5T GE device before and after Gd-TPA administration. In the MR study the following parameters were assessed: tumor dimensions, its parasellar expansion considering the extend of cavernous sinus infiltration according to Knosp's scale, and tumor relation to the optic chiasm. The localization and size of the anterior lobe, presence of nervous part and shape of the pituitary stalk were evaluated. The statistical analysis used the gamma non-parametric correlation test. Results: The study revealed: - strong correlation between the clinical efficacy of neurosurgical treatment and the radicalness of the operation assessed in MRI studies - strong reversed relation between the efficacy of the surgical treatment and the degree of cavernous sinus infiltration observed before surgery. - correlation between the anterior and posterior lobes function and their MR images evaluated prior and respectively after surgery - no correlation between the MR image of the pituitary stalk and the hormonal function of the anterior and posterior lobes, before surgery - correlation between an abnormal shape or deviation of the pituitary stalk and anterior lobe insufficiency as well as posterior lobe function [diabetes insipidus occurrence] after the surgical treatement - close correlation between the field of vision disturbances and the picture of the optic chiasm before and after surgery. Conclusions: MRI of the parasellar region in patients after transsphenoidal pituitary adenomectomy is a highly useful method to evaluate the radicalness of the neurosurgical procedure and to reveal the cause of postoperative hypopituitarism.

361 Precise biochemical classification of Glioblastoma Multiforme tumours by in vivo and ex vivo 1H MR Spectroscopy B. Celda1, D. Monleon1, M. Martinez-Bisbal1,2, J. Piquer3, L. Marti-Bonmati2; 1Química Física, SCSIE, Universitat de Valencia, Burjassot (Valencia), SPAIN, 2Servicio Radiologia, Clinica Quiron, Valencia, SPAIN, 3Servicio Neurocirugia, Hospital-La Ribera, Alzira, SPAIN. Purpose: The aim of this work is to develop a combination of in vivo and ex vivo 1H MR Spectroscopic (MRS) methodology for a precise classification from metabolic composition of Glioblastoma Multiforme (GBM) types, for an accurated diagnosis, prognosis and specific treatment selection of these brain tumours. Subjects and Methods: Spectroscopy studies were done in 10 patients with histopathology diagnosis of GBM. For three of them a

S191

complete biochemical study was done including in vivo MRS and High Resolution Magic Angle (HR-MAS) MRS ex vivo analysis of biopsy tissues. Whereas for the rest of biopsies HR-MAS was applied. In vivo MRS experiments: The MR study was performed in a 1.5 T superconductive unit (Philips Gyroscan Intera). The MRS protocol consisted on single voxel experiments at short (31 ms) and medium (136 ms) TE in the lesion, with TR of 2000 ms, and selective inversion of H2O signal for water suppression protocol. The voxel volume was fitted to the lesional mass. Signals coming from outer volume were suppressed with saturation slabs. HR-MAS of intact tissue, sample preparation and acquisition: The amount of tumoral tissue analyzed for each patient ranged from 20 to 29 mg. The whole HR-MAS study was performed at 0 C. HRMAS spectra were recorded in two Bruker Avance spectrometers at 400 and 500 MHz. Samples were spun at 4kHz and one dimensional spin-echo sequences with water presaturation zgpr and CPMG (Carr-Purcell-Meiboom-Gill pulse sequence [90-(τ-180-τ)n-acquisition] were recorded. CPMG pulse sequence was used as T2 filter. Two-dimensional TOCSY with mixing time of 50 ms and 13C HSQC were acquired with total acquisition time of 15 hours for each sample. Results: The detailed analysis of high resolution HR-MAS spectra of the ten GBM biopsies has allowed the assignment of 127 resonances corresponding to 37 different metabolites. For this extensive biochemical identification a combination of 1D and 2D homo and heteronuclear experiments, including 1H and 13C nuclei, were used. Moreover the standard metabolites a set of different amino and fatty acids were clearly identified. Discussion/Conclusion: ex vivo HR-MAS spectra at 0 C of intact biopsy avoids tissue chemical modifications and degradation. The large number of biochemical compounds identified together with the analysis of different metabolites ratios, as myo-Inositol/Glycine and Phosphocholine/Choline, have allowed a significant tendency for a precise classification of GBM types. These high resolution results can be directly applied to in vivo MRS.

Brain MRS 362 Irradiation induced brain metabolic changes as studied by means of in vitro and in vivo 1H NMR L. Matulewicz1, A. Cichon1, M. Mych2, M. Sokol1; 1Department of Medical Physics, Institute of Oncology, Gliwice, POLAND, 2Department of Medical Physics, University of Silesia, Katowice, POLAND. Introduction: Our study was designed to reveal the impact of radiotherapy on human brain metabolism. The biochemical changes vs. irradiation dose were described by means of a mathematical model. The proposed approach was checked for the irradiated rat brains. Subjects and Methods: 32 patients (mean +/- SD age, 40 +/- 12 years) received a diagnosis of glial tumors and after surgical operation were treated with 3D conformal techniques using 6-20 MV photons. The total dose of 60 Gy was applied in fractions of 2 Gy. The spectra were acquired from the brain regions of different doses and the localization of the volumes of interest was done basing upon the therapy plans. The whole-body MRI/MRS system (Elscint 2T Prestige) operating at a field strength of 2T and a proton resonance frequency of 81.3

S192

Brain MRS

MHz was used equipped with the standard head coil. The spectra were acquired from the volumes of 1.5x1.5x1.5cm3 using a double-spin-echo PRESS sequence with the following parameters: TR=1500 ms, TE=35 ms, 50 acquisitions. The spectra were normalized and post-processed with the automatic fitting in the frequency domain using the residuals method offered by PeakFit 4.0 (by SPSS Inc., Chicago, USA). Spectral lines were approximated under the assumption of the mixed Lorentzian/Gaussian lineshapes. The set of fitted signals was the same for all the spectra under study. It was obtained from the fitting of the 1H MR spectra of the healthy volunteers using the second derivative method. Results: The integral intensities of the signals due to N-acetyl aspartate (NAA), choline based compounds (Cho), creatine and phosphocreatine (Cr) were analysed vs. received dose. The processes influencing the metabolites integral intensities were represented by a sum of Gauss functions and the model was constructed under the assumption of the cyclic variations of the relative contributions of the repair and destructive processes: In the above function A is an oscillation amplitude, λ describes cells reactions on repair and destructive factors, d is an irradiation dose and Dik is a dose value for which maximum is observed. λ is assumed to reflect the disturbance of the cell metabolism.

Conclusion: In response to ionizing radiation cells activate a series of biochemical pathways which results in a oscillative character of the dependencies of the metabolite integral intensities on dose. The model parameters seem to describe the response of the living system for the destructive factor.

363 Cerebral glutamate metabolism of Parkinson disease model rat: an in vivo dynamic 13C MRS study C. Chassain1, G. Bielicki2, J. Donnat2, J. Renou2, F. Durif1; 1Emi 9908, INSERM, Clermont - Ferrand, FRANCE, 2STIM Theix, INRA, Saint Genès Champanelle, FRANCE. The aim of our research is to explore the metabolism of the basal ganglia (BG) in models of Parkinson’s disease (rats with lesion of the nigro-striatal pathway by 6-OH dopamine). 13C MR spectroscopy was used to monitor in vivo glutamine/glutamate synthesis from [2-13C] acetate. The principal interest of these methods is to be able to follow in vivo, the changes in glutamatergic neurotransmission within the BG. We have measured in vivo glutamate concentrations in rat brain according to their physiopathological state and the antiparkinsonian treatment. Study is performed on control rats (n = 6) and on rats with unilateral lesion of the medial forebrain by 6-OHDA (n = 5) in a stable state and after levodopa administration (50mg/kg; i.v.). Rats were anesthetized; [2-13C] sodium acetate is infused in the right jugular vein for 2 hours. 13C NMR spectra are recorded on a Biospec Bruker 47/40 by using a surface probe 1H/13C (30mm in diameter) made in our laboratory. Spectra are acquired in the injured cerebral hemisphere with a volume selection made with 6 saturation bands. The sequence is a 13C acquisition sequence with 1H-decoupling during acquisition. Sequence last for 17 minutes and acquisitions are repeated for 2 hours.

After 51 minutes substrate infusion, the rate of glutamate C4 expressed as percentage of lipids is higher in parkinsonian rats (62 ± 8 %) than in controls (48 ± 12 %; F = 11.37; p<0.01). Glu C4 is also higher after 68 and 85 minutes substrate infusion (62 ± 10 % vs 52 ± 7 %; F = 5.22; p<0.05 and 58 ± 7 % vs 45 ± 5 %; f = 7.81; p<0.05). In parkinsonian rats, levodopa restores Glu C4 rates identical to controls (at 51 minutes: 62 ± 8 % for parkinsonian rats and 44 ± 17 % after levodopa; F = 7.23; p<0.05). Our results show the essential alteration of glutamatergic neurotransmission is a glutamatergic overactivity in Parkinson disease model rat brain. They agree with the current model of the basal ganglia and verify the ex vivo observations performed on tissues from animal models of Parkinson’s disease.

364 1H MRS study of peculiarities of the cerebral and liver metabolism in patients with type 1 diabetes mellitus V. Rogozhyn, Z. Rozhkova; Mri, Radiological Center of the Academy of Medical Sciences of Ukraine, Kyiv, UKRAINE. Purpose: On the basis of the analysis of 1H MRS data we propose the quantitative indicators for the characteristics of the local metabolic state of brain and liver in patients with type 1 diabetes mellitus in chronic as well as acute state. Subjects and Methods: Two groups of patients are studied by MRI and 1H MRS with 1.5T Magnetom Vision (SIEMENS). The 1st group includes 60 diabetic patients in the age from 19 to 65 years without neurological disorders. The 2nd group consists of 75 healthy volunteers in the age from 18 to 73 years. 1H spectra in the brain are recorded in both hemispheres in the occipital lobe (gray matter) and in the frontal lobe (white matter) with the SVS sequence: TR/TE=1365,1500/135,20 ms, VOI = 8cm3, NS = 128. In the liver the spectra are obtained with the STEAM method: TR/TM/TE=600/10/15 ms, and with 2D CSI: TR/TE=5000/20 ms, PN=16x16, NS = 1. Results: In the 1H spectra of the brain the following signals are identified: N-acetylaspartic acid (NAA)-2.02 ppm, total creatine (Cr)-3.03 ppm (CH3 of Cr and phosphocreatine plus minor contributions from γ-amino butyrate and glutathione), choline (Cho)3.24 ppm, glucose (Glx)-3.43 and 3.8 ppm, myo-Inositol (mIns)-3.58 ppm, lactate (Lac)-1.33 ppm and peaks at 2.12-2.22 ppm are given by acetone. In the 1H spectra of liver the signals of the following metabolites are obtained: Cho-3.2 ppm, lipids (Lip)2.0 ppm, glycogen (Glc) - 3.5-4.0 ppm and Lac-1.3 ppm. For the patients of the 1st group the significant decrease of NAA and Cr and the increase of Cho, mIns and Glx peak areas are observed. Presence of Lac is characteristic for the diabetic state in the cases when the blood sugar content is more than 12 mM/L. The signals of Glx and Lac in the spectra of the liver of diabetics are detected more frequently compared to the spectra of volunteers. For the 1st group these observations reflect the loss of capability of the great portion of the liver to regeneration. That leads to the decrease of the oxidative pathway of metabolism and to the increase of glycolysis and accumulation of Lac. Conclusions: 1H MR Spectroscopy gives a new insight into brain and liver biochemistry in chronic as well as acute diabetes mellitus and is very useful for evaluating the disturbances of metabolism of the brain in patients with long-term diabetes.

Brain MRS

S193

365 High-resolution parametric estimation of two-dimensional Magnetic Resonance Spectroscopy D. Belkic; Medical Radiation Physics, Karolinska Institute, Stockholm, SWEDEN. Introduction: Recently, in a clinical application of two-dimensional (2D) MRS, Thomas [1] successfully demonstrated for the first time that 2D Localized Correlated Spectroscopy (L-COSY) can be used to evaluate changes in metabolite ratios. They had two patient cohorts, hepatic encephalopathy and late-life major depression (MDD) whose spectral data were compared with healthy controls. Patients with MDD had a significant decrease in NAA/Cr and increase in mI/Cr, Asp/Cr and Glx/Cr relative to controls. Such findings cohere with data from other studies based upon 1D-MRS. Additionally, these 2D-MRS data show several metabolite changes that have not been recorded previously with 1D-MRS. The 2D-FIDs from Ref. [1] were processed by the 2D-FFT. In the 2D-FFT, good frequency resolution is only possible for long total acquisition times T_1,T_2. Long T_1 could be intolerable for the patient in the scanner. Currently, this limits further clinical applications of 2D-MRS. Moreover, 2D-FFT gives only shapes/contours of 2D-MRS spectra, but not quantification. Hence, alternative processors are needed. The goal of the present work is to introduce the parametric 2D Fast Pade Transform (2D-FPT) [2] which can achieve the frequency resolution for short T_1,T_2 and simultaneously provide accurate quantification for metabolite cross-correlation peaks. Methods/Subjects: The 2D-FPT is defined as the quotient of two bi-variate polynomials at any given frequencies f_1,f_2. This rational function interpolates between the Fourier grids and extrapolates beyond T_1,T_2 by analytical continuation. The Pade polynomials are extracted from the spectrum given by the Green function which is the Taylor series in powers of f_1,f_2 where the expansion coefficients are the encoded 2D-FID. The 2D-FPT and 2D-FFT are identical at the Fourier grids, but only the former method is defined at non-Fourier frequencies. Additionally, 2DFPT can perform quantification by rooting the denominator polynomial. Results: The 2D-FPT has presently been tested on a brain phantom using two 2D-FIDs provided kindly by Albert Thomas who encoded the data on the same GE 1.5T MRI scanner used in Ref. [1]. The computed 2D-spectra are shown in Figs. 1 and 2 at two signal lengths N_1=64,N_1=256, respectively, with N_2=1024 in both cases. Conclusion: The 2D-FPT is introduced for quantification in 2DMRS. Resolution improvements are based upon the extrapolation feature of this parametric estimator. 1. M.A. Thomas et al MAGMA, 15 (Suppl.), 128-129 (2002) 2. Dz. Belkic, MAGMA, 15 (Suppl.) 36-37 (2002); JCMSE, 1, 1-74 (2001)

366 A new quantitative description of the local metabolic brain state from 1H MRS data V. Rogozhyn, Z. Rozhkova; Mri, Radiological Center of the Academy of Medical Sciences of Ukraine, Kyiv, UKRAINE. Purpose: We propose the quantitative indicators for the characteristics of the local metabolic brain state using primary MRS parameters (peak areas) of metabolites: N-acetylaspartic acid (NAA), creatine (Cr) and choline (Cho). Subjects and Methods: Four groups of subjects are examined by MRI and MRS using 1.5 T Magnetom Vision (SIEMENS). The 1st group (VG) includes 70 healthy volunteers (18-73 y). The 2nd group (DG) consists of 60 diabetic patients (19-69 y). The 3rd group (TG) consists of 72 patients (16-79 y) with gliomas (GII, GIII, GIV) and meningiomas. The 4th group (SG) consists of 120 patients after stroke (29-79 y). In the TG the spectra are recorded in

S194

Brain MRS

four locations: in the tumor and intact tissue of the ipsilateral hemisphere, and in the contralateral side on the tumor’s level, and on the level of the examined intact tissue of the affected hemisphere. In the VG, DG and SG the spectra are recorded in two locations: in the occiput and in the forehead. Spectra are recorded with the STEAM sequence: TR/TE=1365/135 ms, VOI=8 cm3. Results: We introduce two indicators: the metabolite content AM as the peak area and the metabolite concentration CM as the ratio of the peak area to the sum of all the peak areas S: CM=AM/S. We describe the metabolic state of the brain by the triad T* = {ACho,ACr,ANAA}, where ACho, ACr, and ANAA are the peak areas of the signals from Cho, Cr and NAA. We believe that each of the areas takes three values: 1, 2 and 3, to give six symbolic spectral configurations: 1* ={1,2,3}, 2* = {2,1,3}, 3* = {1,3,2}, 4* = {3,2,1}, 5* = {3,1,2} and 6* = {2,3,1}. We consider the experimental sets of the content AM and concentration CM as a function of S. Configurations 1* and 2* relate to the normal local brain states. The configurations 3* and 4* are more often in elderly people. Configuration 5* rarely appears in the VG and is relatively often in the DG, SG and TG. Configuration 5* takes place for all of gliomas and sometimes for meningiomas. Configuration 6* appears in SG, in patients with GIV and anaplastic meningiomas. In the TG triads 5* and 6* are observed both for the affected hemisphere, inside and outside the tumor. Conclusion: The main result of this work is establishing a definite empirical connection between the spectral triad configurations and the brain state.

367 Reversible concentration changes of the choline signal in abstinent alcoholics G. Ende1, H. Welzel2, S. Walter1, W. Weber-Fahr1, A. Heinz2, K. Mann2; 1NMR Research in Psychiatry, Central Institute of Mental Health, Mannheim, GERMANY, 2Department of Addictive Behavior and Addiction Medicine, Central Institute of Mental Health, Mannheim, GERMANY. Introduction: This study focuses on metabolic alterations of the cerebellum and the frontal cortex in alcoholics at the beginning of withdrawal and after three months of abstinence in comparison with matched healthy controls detected by use of a proton multislice spectroscopic imaging sequence (MSSI). These investigations are part of a longitudinal study of metabolic changes and their time course in alcoholics before and during withdrawal supported by the Deutsche Forschungsgemeinschaft (DFG). Methods: All 1H MSSI studies were performed on a 1.5 T Siemens Vision system. Two axial planes of 1H MSSI data were collected using a single spin-echo multislice sequence (TE 135/TR 1500). The first 1H MRSI plane was positioned through the cerebellum and pons using anatomical landmarks for reproducible slice positioning and the second slice superior to the lateral ventricles including frontal lobe gray and white matter and anterior cingulate gyrus. Data from two time points of twenty-three patients and eleven controls are presented. Ten of these patients had been completely abstinent whereas thirteen patients continued to drink alcohol. With use of an automated image coregistration and segmentation program (3) all MRSI voxels were corrected for the CSF content as well as the individual point spread function. For evaluation of the repeated study of the same subject both data files were visualized at the same time for selection of voxels from identical positions. Results: No differences were found for NAA signals in patients compared to controls at both measurement times. At the beginning

of alcohol withdrawal we found a significant reduction of the choline signal in the left and right frontal lobe WM, left and right cerebellar cortex, and the cingulate gyrus in the patients compared to controls. Identical voxel positions were chosen from the second MSSI data sets. The Ch signal from the right cerebellar cortex, cerebellar vermis, left and right frontal lobe white matter and from the cingulate gyrus in the abstinent patient group increased significantly whereas it was unchanged in relapsing patients and in healthy controls.

Discussion: The absence of NAA changes implies that there is no measurable neuronal loss or dysfunction in alcoholics. Nevertheless, other studies report a (reversible) NAA decrease in alcoholics (1,2). Our results of reversible choline signal changes support the hypotheses of an altered cerebral metabolism of lipids in membranes or myelin in these patients. We will further investigate the correlation between the Ch signal and reversible atrophy in abstinent alcoholics. References: 1. Seitz D, Widmann U, Seeger U, et al. [1999] Clin Exp. Res., 23:158-163 2. Schweinsburg BC, Taylor MJ, Alhassoon OM, et al. [2001] Alcohol Clin Exp Res 25:924-934 3. Weber-Fahr W, Ende G, Braus DF, et al. [2002] Neuroimage 16:49-60

368 Value of 1H MR spectroscopy for lateralization of temporal lobe epilepsy: a series of 82 patients M. Dezortova1, P. Krsek2, P. Fendrych1, P. Marusic3, J. Hadac4, V. Komarek2, M. Hajek1; 1MR-Unit, Dept. Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague, CZECH REPUBLIC, 2Dept. Paediatric Neurology, Charles University, 2nd Medical School, Prague, CZECH REPUBLIC, 3Dept. Neurology, Charles University, 2nd Medical School, Prague, CZECH REPUBLIC, 4Dept. Paediatric Neurology, Thomayer University Hospital, Prague, CZECH REPUBLIC. Introduction: It has been shown that 1H MR spectroscopy (MRS) could help seizure focus lateralization in temporal lobe epilepsy (TLE), however, it has been rather regarded as a research diagnostic tool. The study was undertaken in order to compare merits of MRS, MRI and EEG in examination of TLE patients. Subjects and Methods: A group of 82 patients from different neurological departments in Czech Republic with clinically identified TLE was examined. Measurements were performed on a Siemens Vision 1.5 T with a CP head coil. MR imaging included standard T2W transversal, T1W sagittal and TIRM coronal slices. MRS protocol consisted of STEAM (TR/TE/TM=5000/10/15 ms, 64 acq) and/or SE (TR/TE=5000/135 ms, 64 acq) sequences of the both hippocampi (volume of interest of 3-5 ml). Coefficient of asymmetry Ca was calculated from the left and right hippocampus for the NAA, Cr and Cho ratios (TE=135 ms). Absolute concentrations of NAA, Cr and Cho were calculated (TE=10 ms) by LCModel and

Brain MRS compared to confirm possible asymmetry. Results of MRS lateralization were compared with MRI and EEG findings. Results: MR imaging showed lesions in the hippocampal area, which could be considered as an epileptogenic focus in 41 patients. MRI pathology was described as mesial temporal sclerosis, atrophy, tumors or cysts. In the case of hippocampal pathology proved by MRI, the lateralization found by MRI, MRS and EEG coincided in 75%. We have found that MRS results correlate especially with invasive EEG seizure focus lateralization. In a significant subgroup of 26 patients, MRS showed pathological hippocampal spectra in MRI-negative patients. Conclusions: Our study clearly demonstrated that MRS can be used as a routine tool for the lateralization of the epileptogenic focus in TLE. It turned out to be more sensitive to some mild pathological changes than MRI. Supported by grants CEZ:L17/98:00023001 and IGA NS 7411-3.

S195

References: 1. Duyn, JH Radiology (1993) 188:277. 2. Soher, BJ Magn Reson Med (1996) 35:356. 3. Vermathen, P Radiology (2000) 214:403. 4. McLean MA NeuroImage (2001) 14:501. 5. Ernst, T. Magn Reson Med (1996) 36:462. 6. Collins, C Magn Reson Imaging (2002) 20:413. 7. Wilson, JL Magn Reson Med (2002) 48:906. Supported by NIH grant RO3AG17364

369 Anterioposterior gradient in choline concentration in the mesial temporal lobe A. Horska, Ph.D., A. Arslanoglu, M.D., P. B. Barker, D.Phil.; Radiology, Johns Hopkins University, Baltimore, MD. Purpose/Introduction: The mesial temporal lobe (MTL) is a common site of age-related and pathological changes. Diagnosis of MTL pathologies may be assisted by proton MR spectroscopic imaging (MRSI). However, MRSI in the MTL is technically challenging, due to poor magnetic field homogeneity in the anterior regions. Our aims were: 1. To assess the feasibility of quantitative MRSI in MTL, 2. To measure metabolite concentrations along long axis of MTL. Subjects and Methods: We examined 20 healthy subjects (20-40 years old, 10M). We also present two cases with suspected pathologies in the anterior MTL (17 year old boy with suspected tumor and 6 year old girl with neurological deterioration). MRSI was performed with TR/TE=2300/280 ms (1). Spectra were evaluated from 5 regions (Figure). Quantitation was performed using the phantom replacement methodology (2) with partial volume corrections for CSF. ANOVA was used to assess regional differences in metabolite concentrations of choline (Cho), creatine (Cr), and N-acetyl aspartate (NAA) and metabolite ratios. Results are presented as averages ± standard deviations, with statistical significance set to p<0.05. Results: Due to insufficient field homogeneity in the anterior MTL, complete data were obtained in 9 controls only. The figure shows examples of spectra, average metabolite concentrations and metabolite ratios. An anterio-posterior gradient in Cho concentration and Cho/Cr was detected along the MTL with the highest averages in anterior MTL (p<0.05 compared to regions 3-5). NAA concentrations and NAA/Cho in the MTL were lower than in the temporal neocortex (p<0.05). No significant differences in Cr concentration were found among the regions. In both patients, high Cho levels in anterior MTL were detected. However, the control spectra did not suggest presence of tumor or specific involvement of MTL, respectively. Discussion/Conclusion: Our results of high Cho in the anterior MTL (uncus) are consistent with prior studies, showing anterior posterior gradients in NAA/Cr+Cho along the hippocampus (3) and high Cho in anterior hippocampal regions (4). Normal high Cho levels should be considered when evaluating MTL pathologies (epilepsy, tumors, Alzheimer’s disease). Several approaches (5-7) may be considered to improve the feasibility of MRSI in MTL.

370 1H MRS choline signal in the prefrontal lobe of male healthy controls: left-right asymmetry and gray/white matter differences G. Ende1, T. Demirakca1, H. Welzel1, S. Walter1, W. Weber-Fahr1, D. Wildgruber2, F. Henn1; 1NMR Research in Psychiatry, Central Institute of Mental Health, Mannheim, GERMANY, 2Department of Neuroradiology, University of Tubingen, Tubingen, GERMANY. Introduction: It is well accepted that the two hemispheres of the brain are structurally and functionally different (1). The frontal lobes are known to be involved in response selection, executive function and emotional processing. Structural and functional asymmetries have been reported by a number of groups (2). In contrast, there is very little evidence of neurochemical laterality. In a recent study by Jayasundar (3) a significant lateralization in the distribution of proton MR visible metabolites for a variety of brain regions has been reported. The presented investigation is part of a multicenter study entitled “ neuronal representation of the communication of emotions” funded by the Heidelberger Akademie der Wissenschaften. Subjects and Methods: All 1H single spin-echo multislice (MSSI) studies were performed on a 1.5 T Siemens Vision (TE 135/TR 1500). The slice was positioned superior to the lateral ventricles including prefrontal lobe grey and white matter and anterior cingulate gyrus. Voxels were selected from the frontal lobe of twelve subjects and were subdivided into “predominantly white matter (PWM)” and “predominantly grey matter (PGM)” voxels with a WM or GM content of more than 50%, respectively. With use of an automated image coregistration and segmentation program (4) all MRSI voxels were corrected for the CSF content as well as point spread function.

S196

Brain MRS

Results: First: the signal of choline containing compounds (Ch) was significantly different in PGM compared to PWM voxels of the same hemisphere (left p = 0.008, right p < 0.0001) emphasizing the need for voxel segmentation and the discrimination of PWM and PGM voxels. The creatine signal from the left hemisphere reached p = 0.03 comparing PWM and PGM. No other significances were found for creatine and NAA. Second: the voxels chosen as PWM showed significant left right differences in their WM and GM content. Nevertheless no significant correlation of the Ch signal and WM or GM content was found. No differences were found for WM or GM content in PGM voxels of both hemispheres. Third: we conducted a regression analysis of the PWM voxel Ch signals for WM and GM content. The resulting residuals for the two hemispheres were still significantly different in a paired t-test (p = 0.05) with higher Ch on the right side. Discussion: The results of this pilot study support the hypotheses that there could be a correlation between functional asymmetries and asymmetry in the concentration of choline containing compounds in the prefrontal lobe. References: 1. Galaburda AM; et al., Science 1978; 199:852-856 2. Wildgruber D et al., Neuroimage 2002; 15:856-869 3. Jayasundar R., Neurol India 2002; Sep;50(3):267-71 4. Weber-Fahr W, et al., 2002; Neuroimage 16:49-60

371 Usability of HMRS in CNS diagnostics of HIV positive patients A. Urbanik1, B. Sobiecka1, J. Kozub1, L. Podsiadlo1, A. Garlicki2, T. Mach2; 1Department of Radiology, Jagiellonian University, Collegium Medicum, Kraków, POLAND, 2Department of Infection Diseases, Jagiellonian University, Collegium Medicum, Kraków, POLAND. Purpose: The aim of the study is presentation of brain metabolic changes of HIV - positive patients in different stages of disease involvement using HMRS. HIV- positive patients with the clinical symptoms of brain injury were compared with the neuroasymptomatic patients. Material and Method: MRI and HMRS were performed in 19 patients (4 women, 15 men), 23 to 49 years old using 1,5 T Signa Horizon (GEMS) unit. Saggital and axial T1 weighted and axial T2- weighted images were performed. The spectral processing was performed using PRESS technigue. MR pictures and spectrum

from the periventrical , frontal and parieto- occipital regions were evaluated. In 4 patients AIDS dementia complex was clinically stated; slowly progressive dementing illness (group 1), while the remaining patients had no clinical changes concerning CNS (group 2). Hemi or tetraplegia, bradypsychia and hypokinesia were observed in the all patients of the group 1. The patient with tetraplegia suffered from the toxoplasmatic encephalitis and maningitis in the past. One patient with hemiplegia was suffered from the right facial nerve paralysis and motor aphasia. Results: MRI: group 1 - brain atrophy and focal hyperintense lesion in T2weighted MR images and hypointense lesion in T1 weighted MR images in the centrum semiovale and in the periventricular white matter, caused by gliosis, were dominating signs. In 1 patient the area of extravasated blood in stage of MHb, after vasculitis was observed. group 2 - in 10 cases no CNS changes were observed, in 5 remaining cases moderate degree of brain atrophy was observed. HMRS: group 1 - significant decease in NAA/Cr and increase in Cho/Cr, moderate increase ml/Cr and Lac/Cr ratios was observed. group 2 - different intensity significant decrease of NAA/Cr ratios, moderate increase in Ch/Cr and significant increase ml/Cr ratios was observed. Conclusions: 1. HMRS is non-invasi, neurochemical method, based on the magnetic resonance phenomenon, that allows the evaluation of the brain metabolism and detection of neuronal markers and in this way expands CNS diagnostic possibilities in HIV positive patients. 2. Stage of involvement of the process and imaging of metabolite changes in cases without morphological manifestation is allowed by HMRS.

372 Dementia diagnosis by 1H-MRS at different localizations and using short and medium Echo Times E. Arana1, M. Martinez-Bisbal2,1, B. Celda2, L. Marti-Bonmati1,3; 1Servicio Radiologia, Clinica Quiron, Valencia, SPAIN, 2Quimica Fisica, Universitat de Valencia, Burjassot (Valencia), SPAIN, 3Servicio Radiologia, Hospital Dr. Peset, Valencia, SPAIN. Purpose: The study of biochemical differences between common Dementias in elderly and Depression by means of MRI and 1H MRS Subjects and Methods: We prospectively studied 64 (45 women and 19 men) patients with cognitive impairment, age 69.5 ± 9.4 years. All subjects underwent laboratory tests,including a chest radiograph, ECG, chemistry profile, CBC count, thyroid function tests, vitamin B-12 and folic acid levels, syphilis serology and structural MR imaging. Major depression and probable Alzheimer's disease diagnoses were made if they fulfilled the inclusion criteria of Diagnostis and Statistical Manual,revision 4 (DSM-IV). Also, probable or possible dementia were tested with the local version of Mini-Mental State Examination (MMSE).The criteria for minimal cognitive impairment: were memory complaint from the patient and a close relative, normal activities of daily living, normal general cognitive functions, and memory performance abnormal for age, but no dementia

Brain MRS according to the DSM IV. Exclusion criteria were structural abnormalities that could produce demential, cortical infarction, tumor subdural hematoma, major sensory loss, confusional syndrome, intoxication, hypovitaminosis, normal pressure hydrocephalus or intoxication. Subjects were not excluded for the presesence of leukoaraiosis. Analysis (MCMB,BC) were unaware of the individuals's clinical details. The studies of 1H MRS were done with Single Volume with TE 31 and 136 ms in parietal grey matter and in temporal lobes. Results: There were 31 patients with Alzheimer's disease (AD), 18 with major depression (MD), 9 presented mild cognitive impairment (MCI) and 6 with vascular dementia (VD). For distinguishing subjects with dementia, the metabolite ratios with higher values on demented subjects were on TE 31 ms, Co/Cr and mI/Cr (Student t-test, p:0.011 and p:0.001), respectively. Conversely, patients with dementia showed lower levels on NAA/Cr with TE 136 ms (Student t-test, p:0.034) and NAA/Co (UMann Whitney, p=0.015) than non demented.When the four diseases were classified, the only ratios which revealed inter-groups differences were Co/Cr, mI/Cr ratios. So, patients with AD presented the highest values of Co/Cr ratios (0.75 ± 0.19 unities, Anova test, p=0.05) and mI/Cr (0.85 ± 0.19 unities, Anova test, p=0.001) on posterior parietal with TE 31 ms compared to other diseases. Conclusion: To a better differentiation and discrimination among dementia and other pathologies and the diagnostic of Alzheimer Disease, two different locations, temporal lobe and parietal grey matter need to be studied. A more precise characterisation also includes the acquisition of at least one short echo time experiment and a middle or long echo time to optimise the results.

373 Effect of gadolium on single voxel MRS of CNS F. Ogresta, J. San Roman, M. Eeleta, D. Elias; Diagnostico por imagenes, Hospital Italiano, Capital federal, ARGENTINA. Purpose: Determine if gadolinium administration changes the Cho/Cr and Naa/Cr ratios during MRS of CNS. Material and Methods: Fifty health volunteers were examined with single voxel CNS MRS with 1.5 T system, with long TE (136 ms). Voxel placement spread over interhemispheric fissure including cortical grey matter of medial aspect of both parietal lobes, immediately behind the corpus callosum. Voxel volume was 25 cubic millimeters. We analyzed Cho/Cr and Na/Cr ratios before and after i.v. administration of gadolinium (0.2 mmol/kg) .Non parametric test (Wilcoxon signed rank test) for match pairs was used for comparision of medians. Significance was assumed if P < 0.05 Results: No significant difference of Cho/Cr and Naa/Cr ratios were found before and after contrast enhanced MRS in this group. Conclusions: Gadolinium could be used with long CNS TE MRS without affecting results. Contrast enhancement could improve location of single voxel MRS over pathologic areas not clearly seen without gadolinium, avoiding heterogeneous areas of lesion.

374 MR imaging of the CNS with gadobenate dimeglumine: Does high relaxivity have advantages? M. Essig1, M. V. Knopp2, V. M. Runge3, G. Pirovano4, M. Kirchin5; 1Radiology, German Cancer Research Center, Heidelberg, GERMANY, 2Radiology, Ohio State University, Columbus, OH, 3Radiology, Scott & White Clinic and Hospital,

S197

Austin, TX, 4Clinical Research, Bracco, Milano, ITALY, 5Medical Writing, Bracco, Milano, ITALY. Purpose: Gadobenate dimeglumine (Gd-BOPTA, MultiHance®) is a gadolinium-based MR contrast agent with high in vivo T1 relaxivity of (9.7 mM-1s-1) compared to other available gadolinium contrast agents due to a capacity for weak and transient interaction with serum albumin. The presentation will summarize the principal results of the clinical CNS studies conducted on Gd-BOPTA in light of the unique protein interacting capability. Materials and Methods: In a first step Gd-BOPTA was evaluated using a cumulative dosing regimen up to 0.3 mmol/kg for MRI of cerebral metastases. Two Phase III controlled trials were conducted to compare Gd-BOPTA at doses up to 0.2 mmol/kg with GdDTPA-BMA (gadodiamide, Omniscan®) at doses up to 0.3 mmol/kg for MRI of patients with CNS lesions. Thereafter, two intra-individual, blinded, fully randomized comparative studies were conducted in patients with either cerebral gliomas or metastases to determine the existence of quantitative and qualitative differences versus Gd-DTPA (gadopentetate dimeglumine, Magnevist®) and Gd-DOTA (gadoteric acic, Dotarem®). For these studies with a dose of 0.1 mmol/kg, quantitative analyses were based on tumor contrast and C/N ratios at ROIs in the tumor, unaffected white matter, and a region outside the head while qualitative assessments were basen on on- and off-site evaluation for lesion contrast, lesion delineation and information upon the internal morphology and structure. Results: In patients with metastases a dose of 0.1 mmol/kg GdBOPTA is sufficient for most clinical situations while a cumulative 0.2 mmol/kg dose provides the best information in most cases.. Comparative studies revealed that 0.1 mmol/kg Gd-BOPTA offers improved performance compared to 0.1 mmol/kg Gd-DTPA-BMA, and that cumulative 0.2 mmol/kg Gd-BOPTA is equivalent to cumulative 0.3 mmol/kg Gd-DTPA-BMA. The intra-individual comparative studies revealed superior qualitative enhancement performance for Gd-BOPTA compared to both Gd-DTPA and GdDOTA for all evaluations (Figure 1). When questioned if they preferred the first or the second contrast agent the on-site investigators named Gd-BOPTA in a significantly greater number of cases (p<0.01). The subjective assessments were confirmed by the objective measurements of signal intensity which showed significantly higher signal intensity and contrast values for Gd-BOPTA (p<0.01). Discussion: Gd-BOPTA is a safe and valuable contrast agent for the assessment of CNS neoplasms. Gd-BOPTA has proven to be superior to Gd-DTPA, Gd-DOTA and Gd-DTPA-BMA in clinical studies conducted to date. This superiority was statistically significant for qualitative subjective assessments and quantitative objective measurements of signal intensity.

S198

Brain MRS

375

377

Normal Pressure Hydrocephalus: Correlation between MR changes & Aqueductal Stroke Volume Y. Ragab1, S. S. A. Lingawi2; 1Radiology, Dr. Erfan & Bagedo Hospital, Jeddah, SAUDI ARABIA, 2Radiology, King Abdulaziz University, Jeddah, SAUDI ARABIA.

The comparison of MR quantitative study of the brain structures in patients with SCA1 and SCA2 R. Poniatowska1, A. Sulek2, E. Zdzienicka2, R. Krawczyk1, E. Pilkowska3, M. Rakowicz4, J. Zaremba2, R. Boguslawska1; 1Magnetic Resonance, Institute of Psychiatry and Neurology, Warsaw, POLAND, 2Genetic, Institute of Psychiatry and Neurology, Warsaw, POLAND, 3Neurology, Institute of Psychiatry and Neurology, Warsaw, POLAND, 4Neurophysiology, Institute of Psychiatry and Neurology, Warsaw, POLAND.

Purpose: To correlate the severity of various brain MR changes in normal pressure hydrocephalus (NPH) with the calculated aqueductal CSF stroke volume. Materials and Methods: The MR Images of twelve clinically suspected NPH patients were retrospectively reviewed. All patients had complete brain MRI studies including phase contrast and quantitative CSF velocity studies across the cerebral aqueduct. The Images were assessed for the presence and the grade of ventricular enlargement, sulcal changes, upward bowing of the corpus callosum, flattening of the gyri against the calvarium, white matter hyperintensities and aqueductal CSF flow void. These changes were graded in a scale of 1 to 3 according to their severity. Results: Nine out of the twelve NPH patients (75%) had exaggerated aqueductal CSF flow void (grades 2 and 3), the aqueductal stroke volume in all these nine patients were increased (greater than 42 micro litter). The other three NPH patients (25%) had either no or grade 1 aqueductal CSF flow void and their aqueductal stroke volume was normal (less than 42 micro litter). The other brain MR changes assessed, did not have a fixed pattern of correlation with the aqueductal stroke volume. Conclusion: In patients with clinically suspected NPH, exaggerated aqueductal CSF flow void correlates positively with increased aqueductal stroke volume.

376 The effect of aging on the apparent diffusion coefficient of normal appearing gray matter Y. Bilgili, B. Unal; Radiology, Kirikkale University, Ankara Turkey, TURKEY. Purpose: The purpose of our study was to study the hypothesis that the apparent diffusion coefficient (ADC) of normal-appearing gray matter increase with advancing age similiar to white matter. Subjects and Methods: We selected 45 patients with normal MR imaging findings from patients undergoing MR imaging. Diffusion-weighted MR imaging was performed and for each patient, the average ADC on trace-weighted diffusion images of the gray matter (at thalamus) were compared with the patient's age. Results: Individual thalamic ADC and patient ages, although showing a trend to higher ADC with increasing age, did not reach statistical significance. Conclusion: Advancing age is associated with a small but statistically insignificant increase of water diffusibility in human gray matter. This may reflect mild structural changes with normal aging.

Introduction: Spinocerebellar ataxia type 1 and 2 belong to the group of neurogenerative disorders, in which progressive cerebellar ataxia occurs. They differ in the frequency of appearance (SCA1 - 60% of all ataxias, SCA2 – 20%) and localization of pathological CAG repeats (SCA1 – chromosome 6p23, SCA2 – 12q24). Purpose: The aim of the study was to assess the differences in the quantitative magnetic resonance study of the patients with SCA1 and SCA2. Material and Method: The MR brain studies of 14 patients with SCA1 and 14 with SCA2 were compared. The SCA1 group consisted of 3 men and 11 women (mean age 42y +/-9), SCA2 group of 10 men and 4 women (mean age 39y, +/- 11). The MR studies were performed using low-tesla unit, with standard SE and FSE sequences with T1- and T2-weighted images in orthogonal planes. On midsagittal T1-weightes image were measured areas of – ventral pons (Pv), dorsal pons (Pd), cerebellum (C), posterior fossa (PF), cerebellum hemispheres (C_h) and peduncles (C_p). All measurements were standardized - expressed as ratios of the anatomy area to the area of posterior fossa. Also the diameter of spinal medulla standardized by diameter of spinal canal was measured on every cervical spine level. Results: The statistically significant difference was found between the measurements in the groups of patients with SCA1 and SCA2. The mean value of Pv was 0,0645 in SCA2 group compared to 0,0863 in SCA1, Pd – SCA2 = 0,0375, Pd-SCA1 = 0,0470; C_h of the left cerebellum hemisphere was 0,05 in SCA2 group and 0,082 in SCA1, of the right hemisphere 0,064 in SCA2 and 0,08 in SCA1; C_p respectively – left and right – SCA2 to SCA1 – 0,32 to 0,42 and 0,35 to 0,43. Also the diameter of medulla on C1 cervical spine level was significantly smaller in SCA2 group compared to SCA1 – 0,34 to 0,42. Conclusion: In both groups – SCA1 and SCA2 the atrophy of pons and cerebellum was found. The pontine atrophy is proportional to cerebellar atrophy and they are both symmetrical. The group of SCA2 patients showed more advanced atrophy of most of the investigated brain structures than the SCA1 group. MR study appears useful diagnostic tool in differential diagnose of these 2 genetic disorders.

Brain MRS 378 Value of MRI in early diagnosis and monitoring of central nervous system lesions (CNS) in patients with AIDS V. Sokolska1, E. Czapiga1, M. Sasiadek1, B. Knysz2, M. Kuliszkiewicz- Janus3, A. Jurczyk3; 1Radiology, Medical University of Wroclaw, Wroclaw, POLAND, 2Infectious Diseases, Medical University of Wroclaw, Wroclaw, POLAND, 3Hematology, Medical University of Wroclaw, Wroclaw, POLAND. Purpose/Introduction: About 10% of all human immunodeficiency virus (HIV) infected and 75- 90% of patients with acquired immunodeficiency syndrome (AIDS) present neurological symptoms and signs. Patients with neurological symptoms are the most difficult diagnostic group among these with AIDS. The purpose of the study was to demonstrate the usefulness of the MR imaging in diagnosis and differentiation of central nervous system (CNS) abnormalities. The spectrum of these abnormalities can be divided into three categories: HIV- associated lesions, opportunistic infections and neoplasms. Subjects and Methods: For MR studies we used EDGE-ECLIPSE 1,5 T with FSE sequences, T1- and T2- weighted images, before and after paramagnetic contrast agent injection, in sagittal, coronal and transverse planes. In almost all cases fluid attenuated inversion recovery (FLAIR) technique was also used. We performed MR examinations of CNS in 30 AIDS patients. 8 of them were examined several times. The radiological abnormalities were compared with pathological findings at laboratory tests, histopathology and autopsy. Results: In 13 cases AIDS dementia complex (ADC) was clinically diagnosed. In these with ADC we observed diffuse cerebral atrophy in 9 patients. It is the most frequent radiological finding. White matter abnormalities were observed in 7 cases, and in 4 patients MR imaging was normal. The abnormalities show no enhancement after application of contrast medium. Toxoplasmosis was recognized in 5 cases, cytomegalovirus infection in 3 cases, progressive multifocal leukoencephalopathy (PML) in 3 patients, tuberculosis in 3 cases , cryptococcosis in 2 patients, candidiasis in 1 patient and primary CNS lymphoma also in 1 patient. Conclusions: The MR technique has a high sensitivity in identify the lesions of the central nervous system. It is a very useful method in difficult cases of abnormal laboratory findings. MR plays an important role in establishing, treatment planning and in therapeutic follow- up evaluation in patient with AIDS. In some cases differential diagnosis in MR is difficult and brain biopsy is still necessary.

S199

sensitive to motion3. We investigated which of both techniques should be preferred for DWI in neonates. Methods: In 14 neonates both SSH and MSH DWI-EPI images were obtained (Philips Gyroscan 1.0T), and in 17 adult patients for comparison. Pulsed Field Gradients with 3 b-values (0, 400 and 800 s/mm2) were used, for single-shot with EPI-factor=101 and TE=94ms (1 minute) and for multi-shot with EPI-factor=13 and TE=22ms (5 minutes), including cardiac triggering and navigator echo for motion detection and phase correction. Both sequences: matrix=128x128, FOV=150mm in neonates, 230mm in adults. Post-processing was performed using Mathematica®. Two radiologists scored quality and diagnosability of both series independently and blinded to clinical diagnosis. The ADC in MSH and SSH in frontal and posterior white matter was compared using the relative ADC difference (RAD). The absolute value of the ADC difference is used because the ADC is randomly larger in SSH or MSH:

Results: In multi-shot, motion artifacts occur more often in neonates than in adults (Table 1). In single-shot, ghosting occurs significantly less often in neonates. Table 1

MSH Adults (17 patients P and 306 slices S)

MSH Neonates (14 patients P and 196 slices S)

SSH Adults (17 patients P and 306 slices S)

SSH Neonates (14 patients P and 196 slices S)

Motion artifact

6 P; 24 S

9 P; 43 S

0 P; 0 S

0 P; 0 S

0 P; 0 S

15 P; 56 S

5 P; 9 S

Distortion 0 P; 0 S artifact

In images without artifacts, RAD is small (Fig 1, group 1). Ghosting evokes significant ADC differences with large standard deviation (group 2). Motion artifacts are not easily recognised in isotropic (averaged) images and also result in large ADC differences (group 3). Both artifacts can easily be mistaken for genuine diffusion abnormalities.

379 Comparison of single-shot and multi-shot DWI-EPI in the neonatal brain C. van Pul1, G. Roos2, S. Derksen2, J. Buijs3, P. Wijn1; 1Applied Physics, Eindhoven University of Technology, Eindhoven, NETHERLANDS, 2Radiology, Maxima Medical Center, Veldhoven, NETHERLANDS, 3Neonatology, Maxima Medical Center, Veldhoven, NETHERLANDS. Introduction: In the neonatal brain it is important to use a fast imaging technique to acquire all Diffusion Weighted (DW)-images needed for ADC calculation. With EPI, the complete k-space can be sampled in one repetition time (single-shot: SSH) or within a number of repetitions (multi-shot: MSH). The first is known to result in considerable distortion1 and N/2-ghost artifacts2 but is less

Multi-shot and single-shot are scored equally with respect to image quality and diagnosability by both radiologists. Discussion: For neonatal brain imaging, less ghosting occurs than is reported in adult brain. Differences between EPI in neonates and adults are likely due to different FOV. For smaller FOV, larger imaging gradients are used, resulting in less distortion and less sensitivity to background gradients. ADC values in multi-shot in neonates are unreliable if motion occurs and therefore single-shot EPI is much better to read-out diffusion information in neonates. References: 1. Haselgrove JC et al. [1996] Magn. Reson. Med. 36:960-964 2. Porter DA et al. [1999] Magn. Reson. Med. 42:385-392 3. Norris DG [2001] J. Magn. Reson. Imaging. 13:486-495

S200

fMRI 380 Neural correlates of velocity discrimination: a functional MRI study in healthy subjects C. S. Wendt1, H. Tost1, M. Ruf1, I. Wolf1, A. Schmitt1, S. Brassen2, D. F. Braus2; 1NMR Research in Psychiatry, Central Institute of Mental Health, Mannheim, GERMANY, 2NeuroImage Nord, Department of Psychiatry, University of Hamburg, Hamburg, GERMANY. Introduction: Although there is only little empirical data concerning the functional neuroanatomy of motion discrimination in healthy subjects, recent studies ascribe the deficient discrimination of small velocity differences in schizophrenic patients to a circumscribed processing deficit in the human homologue of the motionsensitive area MT (hMT, V5). We performed a fMRI study in healthy subjects to clarify the functional role of hMT concerning the discrimination of varying speed velocities. Method/Subjects: 14 healthy volunteers (mean age: 38.6 ± 8.7 years; 11 males, 3 females) performed a block-designed velocity discrimination task which included the pairwise presentation and discrimination of moving sinusoidal gratings with varying velocity differences (easy discrimination tasks: 11°/s vs. 5°/s, difficult discrimination tasks: 8°/s vs. 6°/s). Imaging was performed on a commercial clinical 1.5 T MRI Scanner using a standard EPI sequence, statistical procedures were carried out with BrainVoyager 2000. Results: Velocity discrimination was associated with an enhanced BOLD-response of a highly distributed visuo-motor network (p<0.0001, corrected) entailing the primary and extrastriate visual cortices (V1, V2, V3, V4, BA 17-19) and some higher order areas of the dorsal visual stream in the posterior parietal (SPL, BA 07), right dorsolateral-prefrontal (BA 09), medial frontal and precentral gyrus (BA 06). Furthermore, some brain areas presumably concerned with the complex sequenzing requirements of the task were bilaterally active (cerebellum, anterior insula, basal ganglia). The comparison of the different task conditions yielded a significant enhancement of the higher order processing areas of this network during the difficult discrimination blocks in the posterior parietal cortex (right side: SPL BA 07, left side: IPL BA 40), the right frontal eye field (MFG BA 06), the right dorsolateral-prefrontal cortex (BA 09) and the right anterior insula. Conclusions: The present data suggest that the active discrimination of large and small velocity differences requires the integrated action of a broad visuo-motor network. The task-related activation focus in prefrontal regions may probably originate from the comparably high task loading on elaborate cognitive functions like working memory and response flexibility. In contrast to higher order processing areas of the dorsal network, our data do not support the notion of an extraordinary functional role of area hMT concerning the discrimination of velocity differences. Therefore, the functional anatomy of the observed motion discrimination deficits in schizophrenic patients should be investigated thoroughly with imaging methods (supported by DFG BR 1112/5-1).

fMRI 381 The influence of design factors and fMRI analysis methods assessed by ANOVA E. Dimitriadou1, M. Barth2, K. Hornik1, E. Moser3,2; 1Statistics, University of Technology, Vienna, AUSTRIA, 2Dept. of Radiodiagnostics, University and General Hospital, Vienna, AUSTRIA, 3Dept. of Medical Physics, University, Vienna, AUSTRIA. Introduction: Exploratory data analysis (EDA) methods are popular in fMRI analysis, however, the influence of data set characteristics as well as of various EDA parameters is not clear. In our study, we varied several design factors (2 noise characteristics (NC), 3 CNR levels, 3 initial number of clusters (INC)) and calculated two performance coefficients to validate the influence of these factors and their levels on eight clustering methods [1]. Subjects and Methods: Performance assessment was done using the weighted Jacard coefficient (wJC) [2], which is defined as JC=a/(a+b+c) [a=#TPs, b=#FNs, c=#FPs], adding the inverse of the probabilities of a TP, FN, and FP as weights to account for the small number of TPs in fMRI sets. We also calculate the correlation (CC) between the center of the activation cluster found and the reference center. CC reflects the quality of the cluster center, while wJC provides a more quantitative measure which emphasizes TP scores. Used methods [1] included k-means (km), maximin-distance (mm), hard-competitive-learning (hardcl), neural-gas (ng), cmeans (cm), unsupervised-fuzzy-competitive-learning (ufcl), self-organizing-maps (som), clara (clara) with 50 repetitions each to account for instabilities by random initializations. Artificially constructed data sets (gaussian noise and in-vivo noise) with known activation were used with CNR levels of 1.33, 1.66, and 2. The significance of factor level variations and different methods was tested against the the mean result of wJC (CC) over all factor levels and methods (table:"Intercept") using ANOVA without twoway interactions. Results: The results (multiplied by 100 for readability) are given in the tables. INC is the factor with the most dramatic impact on method performance, decreasing the quality of the result with fewer clusters (reduction of about 40% for wJC; 20% for CC). As expected, CNR is important too (20% reduction from high to low CNR for wJC; 14% for CC). NC has the lowest impact (6% reduction for wJC from artificial to hybrid sets; 14% for CC). Discussion: The results support the assumption that changes in INC, CNR and NC have an important impact on the performance of the algorithms, and, in particular, cluster initialization is very crucial. The results also show that k-means and neural-gas are the best performing EDA methods on average. Acknowledgements: Supported by the Austrian National Bank Fund (OeNB-Project-No.9201). References: 1. M.Barth et al, ISMRM (2003) 2. M.Anderberg, p. 89&125 (1973).

fMRI ANOVA Results (Methods) Effect

Est. (wJC)

Sign.

Est. (CC)

Sign.

Intercept

56.21

km

12.40

***

18.58

63.18 ***

mm

-11.24

***

-22.69

***

hardcl

-1.33

n.s.

9.11

***

ng

11.61

***

19.73

***

cm

1.97

n.s.

-11.19

***

ufcl

-15.06

***

-7.63

***

som

0.97

n.s.

2.97

***

clara

1.32

n.s.

-17.09

***

S201 quency of acoustic hallucinations of approx. 50%, while the loudness of the hallucinations remained unchanged over the 4 weeks of stimulation. FMRI demonstrated a BOLD-Effect activation reduction after TMS in speech related areas, which exceeded the local stimulation area. Discussion: FMRI-data revealed an activation reduction in temporal and temporoparietal areas after TMS corresponding to the clinical recovery. The combination of TMS and functional imaging is promising, allowing an insight into neuro-biological mechanism during TMS intervention, which may help to improve technique and treatment success.

***: <0.0001, **: <0.001, *: <0.05, n.s.: not significant

383

ANOVA Results (INC, CNR, NC) Effect

Est. (wJC)

Sign.

Est. (CC)

Sign.

Intercept

56.21

INC: 5

-23.87

***

-13.49

***

INC: 10

4.46

***

2.18

***

INC: 20

19.42

***

11.30

***

CNR: 1.33

-9.48

***

-7.71

***

CNR: 1.66

1.10

n.s.

1.50

*

CNR: 2.00

8.39

***

6.21

***

NC: artificial

2.68

***

7.18

***

NC: hybrid

-2.68

***

-7.18

***

63.18

***: <0.0001, **: <0.001, *: <0.05, n.s.: not significant

382 Functional magnetic resonance imaging (fMRI) for therapeutic monitoring of transcranial magnetic stimulation (TMS) F. L. Giesel1, A. Hempel2, E. Hempel3, T. Wuestenberg4, U. Seidl5, J. Schroeder2, M. Essig1; 1Department of Radiology, German Cancer Research Center, Heidelberg, GERMANY, 2Department of Psychiatry, University of Heidelberg, Heidelberg, GERMANY, 3Department of Medical Engineering, Forschungszentrum Karlsruhe, Karslruhe, GERMANY, 4Department of Neurology, Humboldt-University of Berlin, Berlin, GERMANY, 5Department of Psychiatry, Universtiy of Heidelberg, Heidelberg, GERMANY. Introduction: Recent studies demonstrated acoustic hallucinations in patients with schizophrenia to be associated with activations in the superior temporal lobes. Based on the hypothesis of focal cortical inhibition low-frequency transcranial magnetic stimulation (TMS) was used, resulting in a slight reduction of hallucinations in some patients. In this case study functional magnetic resonance imaging was employed to image BOLD-effect changes in the temporal lobes under TMS-therapy. Methods: In a curative attempt, a 30-year old schizophrenic patient (DSM-IV) with medication resistent acoustic hallucinations was treated with low-frequency TMS (fstim=1Hz) over a four week period. The TMS-effects were detected based on the auditory hallucinations rating scale. FMRI was performed in a 1.5 T clinical scanner (Magnetom Vision plus, Siemens, Erlangen, Germany) using the standard head coil and a GE-EPI sequence (volume of 30 slices, FOV 240 x 240 mm2, voxel size = 1,88 x 1,88 x 4 mm3, flip angle 90°, TR= 4.7 ms, TE 54 ms) using a design containing acoustic hallucinations. FMRI was performed prior to and after the TMS series to visualize possible cortical activation changes in the stimulated area. Data analyses were performed with SPM99 (http://fil.ion.ucl.ac.uk.spm). Activated voxels were identified by the General Linear Model approach for each condition. Results: After the third week, the patient presented a reduced fre-

An educational tool for a better understanding of k-space through the reconstruction of magnetic resonance images D. Moratal-Pérez1, L. Martí-Bonmatí2, M. E. Brummer3, J. Millet-Roig1, F. Castells1, J. J. Rieta1; 1Ingeniería Electrónica, Universitat Politècnica de València, Valencia, SPAIN, 2Radiology Department, Hospital Universitari Dr. Peset, Valencia, SPAIN, 3Radiology Department, Emory University School of Medicine, Atlanta, GA. Introduction: In magnetic resonance imaging (MRI) the user has the control over how the data are acquired and how they can be manipulated in order to show the reconstructed image. Adjusting several parameters, the user can modify the spatial and temporal resolution, the field of view, the contrast, the speed of the acquisition, the influence of multiple artifacts and several other parameters that will contribute to create the final image. The agent that makes this possible is known as k-space and it refers to the data matrix obtained directly from the magnetic resonance scanner before any kind of processing and before the application of the Fourier Transform, which will provide us of the reconstructed final image1,2. Materials and Methods: This tool has been developed using MATLAB R6.1 (Mathworks, Inc, Natick, MA). It has been created with a Graphical User Interface, in order to facilitate its use. Results: The educational software tool that is introduced here tries to show in an intuitive and didactic way what happens to the reconstructed image associated to a k-space to which some basic processes –like low, high and band-pass filtering among othershave been applied. It offers also the possibility of adding noise or spikes in order to study the behaviour of the k-space and its associated image. It is also possible, within this education tool, to learn how some basic reduced Field-of-View techniques like Rectangular Field-ofView and Half Fourier Imaging work, applying them to the k-space and observing how it is modified and so its associated image does. Discussion/Conclusion: This tool has been tested by radiologists achieving a high level of satisfaction and accomplishing its main Objective: to help to better understand the unknown concept of k-space and how the image is affected by modifying it. It is interesting to remark that this educational software keeps in continuous growth. References: 1. R. Mezrich, Radiology 1995, 195:297-315 2. D.B. Twieg, Med Phys 1983, 10(5):610-621

S202

fMRI quantification, visualization, communication of the findings, and archiving. A more automated workflow could be achieved by minimizing incompatibility issues due to propriertory image file formats and by introducing self written software for quality control and database managment. Vendor and modality independent processing was possible. Discussion/Conclusion: This generic approach facilitated advanced image analysis and allows an effective integration of visualization and analysis tools into a PACS-oriented, radiological environment to improve research capabilities as well as clinical service.

385

Figure 1. Main window of this k-space tutorial. At the upper part of the window we can see the k-space and its image associated to which all the desired operations will be applied. After having applied the desired operations, it is also possible to save the resulting image and its k-space.

384 Current concepts and applications for advanced image processing in MRI H. von Tengg-Kobligk1, K. T. Baudendistel1, J. T. Heverhagen1, M. J. McAuliffe2, T. Kurc3, J. H. Saltz3, M. V. Knopp1; 1Department of Radiology, The Ohio State University, Columbus, OH, 2Center for Information Technology, National Institutes of Health, Bethesda, MD, 3Department of Biomedical Informatics, The Ohio State University, Columbus, OH. Purpose/Introduction: Advances in recent development of post processing equipment facilitate more elaborate image analysis on large data sets acquired using ultrafast MR, multi-slice CT, time dependent imaging data from these modalities, as well as ultrasound and nuclear medicine/PET imaging. The current PAC-Systems are not able to carry out the required functional image analysis. Most available analysis applications are being developed for a single specific purpose such as MR angiography, virtual colonoscopy, etc. In addition, radiological imaging is increasingly utilized for functional analysis and therapy monitoring applications requiring the assessment of multiple imaging studies. Currently, most clinical research groups focus on developing targeted applications for their specific needs without developing a generic infrastructure framework which then may be utilized for multiple modalities and applications. Subjects and Methods: We present an implementation of a general image processing framework for MR image analysis and for the integration of other imaging modalities. The implemented heterogeneous PCs/Linux-based computing cluster uses modules developed in the programming languages IDL (Research Systems, Inc., Boulder, USA) and JAVA (Sun Microsystems, Inc., CA, USA). Developed applications are executed as pre-compiled run-timemodules which then can interface with third party imaging devices and workstations. Large scale computational image assessment for follow-up studies and computer-aided diagnosis was enabled by grid-enabled cluster computing capabilities. Results: We identified, that all specific solutions have the following key components in common: image retrieval, image information database management, human interface, processing,

White matter differences in autism: a voxel based morphometric study G. D. Waiter1, J. H. G. Williams2, A. D. Murray3; 1Cognitive Neuroimaging Research Unit, School of Psychology, Mathematics and Computing Science, University of Aberdeen, Aberdeen, UNITED KINGDOM, 2Department of Child Health, University of Aberdeen, Aberdeen, UNITED KINGDOM, 3Department of Radiology, University of Aberdeen, Aberdeen, UNITED KINGDOM. Introduction: Autism(1) is a developmental disorder involving a wide variety of behavioural and cognitive abnormalities, with characteristic impairments in social communication. There is evidence for a genetic link(2) and also that pathologic mechanisms are in effect during the early stages of brain development(3). These factors suggest that structural differences may be present in the brain of subjects presenting with autism. This structural MRI study was conceived to test this hypothesis by imaging a group of young male adults with Autism with an age, IQ and handedness matched normal control group. Methods: 16 young men diagnosed with Autism using ADI- revised (2000) and ADOS-G, were imaged. They were matched to 16 normal volunteers on age and IQ measured using either the WAISR or the WISC-R. Imaging was performed using a 1.5 T scanner (CVi, General Electric Medical Systems, Milwaukee, WI). A quadrature head coil was used to obtain high-resolution images using an SPGR sequence: FOV, 24cm; 20/6, (TR/TE); flip angle, 35o; slices, 124; slice thickness, 1.6mm; matrix, 256x192; and in-plane resolution, 1 x 1 mm. A T1 template image was generated based on the average of all 32 participants. Preprocessing was based on the “Optimized VBM Protocol”(4) using SPM, with normalisation to the new group specific T1 template. Regionally specific differences in white matter between groups were assessed with an analysis of co-variance, with total brain volume as a nuisance variable. The resulting t-statistics were thresholded at p=0.01 (uncorrected) for display, and maxima at p<0.005 (uncorrected) are reported. Results: Significant reduction in white matter volume was found throughout the brain with particular focus in the middle and posterior regions of the corpus callosum (CC), Fig. 1. No significant areas of increased white matter volume were found.

fMRI

S203 time) , the activated volumes of both sensorimotor cortex increased. The activated volume in lesion side sensorimotor cortex was more than that in non-lesion side sensorimotor cortex. The other patient (MRC Scale: IV) was observed extended activation in lesion side sensorimotor cortex. Conclusion: fMRI allows study of motor function in deep cortical infarction. We were able to investigate differences in motor activation according to the individual MRC Scales. fMRI may be useful tool to monitor and study deep cortical infarction and may be important in understanding of the function of deep cortical area.

387

Figure 1. Regions in which ASD subjects show significant decreases in white matter compared with controls (p<0.01) Conclusions: We have shown that significant local differences in white matter volume are present when comparing autistic subjects with normal controls. Reduced CC size, as reported here, has been found in a number of studies(5,6). The CC is the major axonal pathway linking the two hemispheres, and abnormalities in the CC may therefore result in anomalous connectivity of the cerebral cortex. References: 1. Kanner L. Nerv Child, 2;217-50:(1943) 2. Bailey A.. Psychol. Med. 25;63-77:(1995) 3. Courchesne E. Curr. Opin. Neurobiol, 7;269-278:(1997) 4. Good C. NeuroImage, 14;21-36:(2001) 5. Egaas B. Arch. Neurol. 52;794-801:(1995) 6. Piven J. Am. J. Psychiatry 154;1051-1056:(1997)

386 Motor fMRI in acute deep cortical infarction G. Chung, Y. Han, S. Lee, S. Jeong; Diagnostic Radiology, Chonbuk National University Hospital, Jeonju, Jeonbuk, REPUBLIC OF KOREA. Purpose: To assess cerebral activation of motor function after deep cortical (lentiform nucleus and thalamus) infarction Materials and Methods: We studied motor function of eight righthanded deep cortical infarcted patients (mean age, 61 years; 7 men and 1 woman) who suffered a single unilateral deep cortical (lentiform nucleus or thalamus) infarction. The grade of muscle power by the Medical Research Council were II (two patients), III (three patients), and IV (three patients). All MRI experiments were performed with a 1.5T scanner. The fMRI protocol consisted of eight alternating periods of task performance and rest. The activation tasks consisted of both finger movements. Data analysis of activated area, and calculation of activated volumes in sensorimotor cortex. Results: Six lentiform nucleus acute infarction patients: One right hemiparetic patient (MRC Scale: II), only right sensorimotor cortex (unilateral non-lesion side) were activated. In five (MRC Scale: III-IV) of six lentiform nucleus infarcted patients, bilateral activation of the primary sensorimotor cortex and supplementary motor area was recorded. In four of five bilateral activated patients, extended activation in lesion side sensorimotor cortex was observed. Two right thalamic infarction patients: bilateral activation of primary sensorimotor cortex was recorded. One patient (MRC Scale: II) was observed extended activation in non-lesion side sensorimotor cortex. Follow up fMRI after 40 days (MRC Scale: III at that

FMRI in estimation of influence of difficulty of a cognitive task on the patterns of brain activity A. Urbanik1, M. Binder2, B. Sobiecka1, J. Kozub1; 1Department of Radiology, Jagiellonian University, Collegium Medicum, Kraków, POLAND, 2Department of Psychophysiology, Jagiellonian University, Collegium Medicum, Kraków, POLAND. Purpose: The aim of the study was to examine the effects of the difficulty of the cognitive task on the changes of the patterns of the brain activity. We also determined whether the higher level of difficulty affects in the same way task based on verbal and nonverbal material. Materials and Methods: Twelve volunteers were examined in the Signa Horizon 1.5T MR (BOLD). Experimental design consisted of four activation runs. Each of them utilized another stimulus and the level of difficulty. During the first and the second run subjects were required to remember the changing sequences of letters and the length of the sequences was different for the easy and the difficult task. In the third and the fourth run subjects were asked to remember changing sequences of abstract pictures, the number of pictures was different depending on the difficulty of the task. Statistical analyses were carried out using SPM99 software. Results: The most significant increase associated with difficulty was observed in the dorsolateral prefrontal areas, both in verbal and nonverbal tasks. Furthermore, there were separate activation sites for the nonverbal (parietal cortex) and verbal tasks (precentral gyrus, left parietal cortex), not sensitive to the task difficulty. Conclusions: Areas responsible for various aspects of cognitive function were differently activated by the level of cognitive demand. Regions responsible for manipulation showed increase of activity, which was dependent of the level of difficulty. This was not observed for the areas whose activity was dependent of the type of the task (i.e. storage of verbal and nonverbal material).

388 Comparison of the response of Finger Tapping (Active) and Finger Stimulation (Passive) in fMRI Y. Kim1, Y. Ryu1, B. Choi2, S. Chung3, C. Oh1; 1Biomedical Engineering, Korea University, Seoul, REPUBLIC OF KOREA, 2Biomedical Engineering, Catholic Univ., Seoul, REPUBLIC OF KOREA, 3Reaearch Center, Medinus Co., Yongin-Si, REPUBLIC OF KOREA. Introduction: Two fMRI experiments are performed and the results are compared, one with finger-tapping and the other with finger sensory stimulation. In fMRI with stimulation, a set of images are acquired with some stimulation according to a given paradigm. For motor cortex areas, fMRI experiments with finger tapping are successfully performed to show the activation. From the experi-

S204 ments we have found and compared the common as well as different areas for the areas excited by finger tapping and those by finger sensory stimulation. Methods: We have performed two kinds of experiments; one was done with right finger tapping and the other with right finger stimulation by rubbing with a soft material. Four volunteers aged 23 – 29 (1 female and 3 male) participated in the inter-subject study. The experiments were performed using a 3.0T MRI system (Magnum 3.0T, Medinus Co.) at CMC (Catholic Medical Center in Seoul, Korea). For fMRI measurements, the BOLD contrast method using a single shot gradient-echo echo-planar imaging (EPI) sequence was employed. The parameters were, TR 2400 ms, flip angle 90 degrees, FOV 24 cm, data acquisition matrix 64 by 64 and slice thickness 6mm. Functional studies were performed using a sequential right finger tapping task and rubbing volunteer’s fingers. All the experiments have been performed according to the paradigm in Fig. 3 Results: Figures 1 and 2 show the activated area mapped onto the reference image obtained by using SE sequence. Figure 1 shows the areas activated by the finger tapping and Fig. 2 shows the area activated by rubbing finger. According to the experimental results, there are common areas activated by right finger tapping task and by the finger rubbing. Figure 3 shows the BOLD signal changes in the activated area.

Discussions: As seen in the experimental results, there are common as well as different areas between activated area by right finger tapping task and activated area by the stimulation of rubbing right finger. And there are more activation in the brain during the finger tapping task than receiving the sensory stimulation. According to the results, we can expect that we will be able to discover the differences between the subtle brain activities related to action and sensing of human body using fMRI. References: 1. Toshiharu Nakai, et al. Post-stimulus response in hemodynamics observed by functional magnetic resonance imagingdifference between the primary sensorimotor area and the supplementary motor area. Magn Reson Imaging 2000; 18:1215-1219

fMRI 389 Using Sensitivity Encoding (SENSE) at 3 Tesla for fMRI of brain areas prone to susceptibility artefacts S. Boujraf1, P. Summers1, D. Weniger2, S. Kollias1; 1Institut für Neuroradiologie, UniversitätsSpital Zürich, Zürich, SWITZERLAND, 2Neurology Clinic, UniversitätsSpital Zürich, Zürich, SWITZERLAND. Purpose: BOLD contrast is sensitive to image resolution, relaxation-times, static magnetic field strength and the size, orientation and distribution of blood vessels in the tissues imaged. While increasing field strength brings a gain in BOLD effect, it is also accompanied by greater image distortion due to increased susceptibility effects. A language task was used to test the use of SENSE (1) to facilitate fMRI of functional brain areas near the skull base on a 3 Tesla scanner. Materials and Methods: 8 healthy volunteers underwent wholebrain fMRI at 3 Tesla (Philips Intera) using a single-shot, gradientecho, EPI sequence (4 mm thick, 0.4 mm slice gap, FOV 22 cm, 1282 recon matrix, TE/TR 40/2380 ms, flip angle 90°). Imaging was repeated using SENSE to reduce the number of phase encoding steps by a factor of 2.75. The task consisted of silent word generation alternating with counting in 30 second blocks. Scanning lasted 3.0 minutes yielding 72 volumes per study. The functional test was performed twice for each of the above data acqusition strategies. Data underwent identical post-processing (2), with only statistically significantly correlated (P<0.001) pixels considered as activated. Results: Susceptibility and quadrature ghost artifacts were notably less prominent in the images acquired with SENSE. This made possible visualization and functional mapping of areas in the skull base obscured by artifacts in the conventional EPI studies. This is shown (Figure 1, top row) by left fronto-opercular region (Broca’s area) activation only in the SENSE images. A second observation is that the total number of activated pixels, and detected activation in nonprimary areas are notably reduced in the SENSE images. This is seen in the absence of activation in the right fronto-opercular region (contralateral Broca’s area) and reduced activated volume in the left hemisphere (Figure 1 lower row). Discussion and Conclusion: Using SENSE significantly reduced ghost and skull-base susceptibility artifacts in the fMRI studies at 3T. Unfortunately, a reduction in SNR accompanies the use of SENSE reducing sensitivity to BOLD effects. Together with reduced blurring, this may explain the decrease in detected area of activation seen in the SENSE images. We note that the SNR reduction seen here is greater than expected by SENSE alone and are investigating the software control of receiver bandwidth as a possible contributory factor. References: 1. MRM 1999,42:952-962 2. MRM 1993,30:161-173 Table 1. Comparison of SENSE and non-SENSE image Characteristics

Percentage BOLD signal change

Images acquired with SENSE

Images acquired without SENSE

Ratio of SENSE : non-SENSE results

4.52±0.66

4.01±0.03

1.12±0.13

422.5±6

0.51±0.03

Signal to noise ratio 216.5±3.6

Spectroscopy: Sequences and Techniques

S205

Conclusions: Very short echo times about 10 ms can be achieved compared to TE=25-30 ms reported with other schemes. This enhances coupled resonances such as Glu and Gln due to minimization of phase distorsion of their multiplets. The STEAM excitation scheme is known to give less contaminating signals than the PRESS technique in the short echo time regime, thus avoiding any outer volume suppression and special data processing steps. The inherently lower signal provided by the STEAM excitation is at least partially compensated by less T2 signal decay compared to the previous protocols. The gain in resolution is achieved due to improved field homogeneity across the extremely small voxels and subsequent frequency and phase corrected averaging. References: 1. Gruber S, Mlynárik V, Moser E [2003] Magn.Reson.Med. 49:299-306. 2. Wiedermann D, Schuff N, Matson GB et al. [2001] Magn.Reson.Imaging 19:1073-1080 3. Chard DT, McLean MA, Parker GJM et al. [2002] J.Magn.Reson.Imaging 15:219-225

Spectroscopy: Sequences and Techniques 390 Anatomically matched short-echo time spectroscopy of human brain at 3T V. Mlynárik1, S. Gruber1, A. Stadlbauer1, Z. Starcuk2, E. Moser1; 1Dept. of Medical Physics, University of Vienna, Vienna, AUSTRIA, 2Inst. of Scientific Instruments, Academy of Sciences of the Czech Republic, Brno, CZECH REPUBLIC. Introduction: High resolution proton spectroscopic imaging is a useful tool for mapping metabolite concentrations in relatively small cerebral structures. Recently, a long echo-time (TE=136 ms) high resolution experiment has been reported which, however, provides information on NAA, Cr, Cho and Lac concentrations only [1]. The short-TE experiment would enable to quantify additional metabolites such as myo-Ins or Glx. However, short-TE protocols usually give spectra contaminated by lipid signals from the skull and by water signal from regions of high susceptibility inhomogeneity. In contrast to previously described short-TE protocols [2,3] we used a STEAM-excitation scheme. Subject and Methods: Five healthy volunteers were examined on a 3T Bruker MEDSPEC system. A 3D spectroscopic imaging sequence was used with an excited (TR/TE/TM=1600/11/60 ms) volume of approx. 7x8x3 cm3, which was phase (2D) and Hadamard (1D) encoded in 24x24x4 steps (the actual voxel size of 0.33 cm3). The data were zero filled to 32x32x4 voxels. Results: Fig. 1 shows a morphogical image of one slice in the brain with the overlaid grid of spectroscopic voxels. The spectra extracted are sums of marked voxels in white matter at the level of cella media and in occipital grey matter. Quantitation of myo-Ins and Glu/Gln from these spectra is feasible.

Fig. 1

391 Recovering the lactate signal at 3T T. Lange, U. Dydak, R. F. Schulte, A. H. Trabesinger, P. Boesiger; Institute for Biomedical Engineering, University and ETH Zurich, Zurich, SWITZERLAND. Introduction: Lactate plays an important role in many brain pathologies. Since in vivo concentrations are usually quite low, a high signal-to-noise ratio is required for detection. However, due to “anomalous J-modulation” [1], resonances of coupled spin systems like lactate suffer from signal cancellation in spectroscopy methods using a localization technique such as PRESS or STEAM. Chemical shift displacement artifacts for the refocusing pulses result in a different coupling evolution for different spatial regions of the excited spectroscopic imaging (SI) slice, such that spatial interference effects can lead to a considerable loss in signal intensity for echo times near odd multiples of 1/J (J being the coupling constant). It is shown here that, at a higher field strength (3T), where the chemical shift difference between the coupled spins is comparable to the radiofrequency (RF) pulse bandwidth, a severe underestimation of lactate can occur and how this can be prevented in SI experiments.

S206

Spectroscopy: Sequences and Techniques

Methods: All experiments were performed on a 3T Philips Intera whole body system using a send-receive head coil. 2D SI data (16x16) were acquired with an echo time of 144 ms (=1/J) from a lactate phantom and from the brain of a visually stimulated migraine patient, who gave his written informed consent for this study. To estimate the signal loss due to anomalous J-modulation an SI sequence containing one slice-selective excitation pulse and one refocusing pulse was used. The lactate signal was evaluated for acquisitions with the refocusing pulse being slice-selective versus spatially non-selective. Results: Figure 1 shows the anatomical scout images of the phantom measurement (Fig. 1a) and the in vivo measurement (Fig. 1d). For both cases, a spectrum from one SI voxel is shown, acquired once with a slice-selective (Fig. 1b and 1e) and once with a spatially non-selective (Fig. 1c and 1f) refocusing pulse. The comparison demonstrates the considerable signal loss of lactate, in vitro as well as in vivo. In vitro the signal loss turned out to be 63.3 %, which is in good agreement with theory (65.9 %). Discussion: It was shown that signal loss due to anomalous J-modulation is avoided by performing SI without spatially selective refocusing pulses. In particular the use of PRESS volume selection in combination with SI experiments is therefore discouraged. However, especially in vivo, a good outer volume suppression technique is required to compensate for the missing volume selection. References: 1. Kelley et al. [1999] JMRI 9:732-737. Figure 1

ton MRS at 3.0T. The higher field strength is expected to improve spectral resolution and SNR. Subjects and Methods: The measurements were performed in healthy volunteers on a 3.0T Intera whole body MR system (Philips Medical Systems, Best, The Netherlands). A mid-ventricular shortaxis cine was acquired to determine the trigger delay for end-systole. A B0-map at the same position was measured for calculation of localized 2nd order shim values and resonance frequency. The linear fine shimming and water suppression calibration as well as the spectra acquisition was double triggered: The PRESS excitation was performed at the determined trigger delay after the R-wave each time the respiratory displacement was within the gating window of 5mm around end-expiration measured with a navigator. Parameters: voxel size: 10x18x22mm3, echo time: 41ms, bandwidth: 2000Hz, data points: 1024, repetition time: at least 2900ms, averages: 64. Spectra were averaged, phased, zero filled to 2048 data points and exponentially filtered with 3Hz. Results: Figure 1 shows a spectrum from a 4ml volume in the septum. Peaks from methylene protons (Cr2) of creatine/phosphocreatine (Cr), trimethylammonium compounds (TMA), methyl protons (Cr3) of Cr, and intra- and extramyocellular lipids (IMCL, EMCL) can nicely be observed.

392 Cardiac single voxel proton spectroscopy at 3.0T M. Schär, S. Kozerke, P. Boesiger; Institute for Biomedical Engineering, University and ETH Zurich, Zürich, SWITZERLAND. Introduction: 1H MR Spectroscopy (MRS) allows investigating the lipid and creatine metabolism non-invasively in the human heart. The observable lipid may be a marker of myocardial viability [1] and the total creatine content is reduced in the failing heart [2]. Cardiac MRS is challenging because the underlying motion causes displacements on the order of the localized voxel and difficulties in shimming and water suppression calibration. Respiratory and cardiac double-triggering based on the ECG signal [3] has proven to provide improved spectral quality and reproducibility. Navigator based gating and volume tracking for cardiac MRS [4] has shown increased visibility of creatine and lipid components invivo. Up to now, cardiac proton MRS has been performed at 1.5T only. The goal of this work was to show feasibility of cardiac pro-

Discussion: The feasibility of cardiac proton spectroscopy at 3.0T has been demonstrated. The peaks of interest seem to be better discriminable compared to spectra measured at 1.5T. Reproducibility and a direct comparison between 1.5T and 3.0T have yet to be done. References: 1. den Hollander, JA, et al. MRM 32:175-80, 1994 2. Bottomley, PA, et al. Lancet 351:714-18, 1998 3. Felblinger, J, et al. MRM 42:903-910, 1999 4. Kozerke, S, et al. Proc ISMRM, p. 1698, 2002

Spectroscopy: Sequences and Techniques

S207

393

394

Spurious signals in DQF spectroscopy: diffusion-attenuated two-shot stimulated echoes J. Pictet1, J. J. van der Klink1, R. Meuli2; 1Faculté des Sciences de Base, EPFL, Lausanne, SWITZERLAND, 2Diagnostic and Interventional Radiology Dept., CHUV, Lausanne, SWITZERLAND.

Observations on J-coupled spins in SSFP based 1H Spectroscopic Imaging C. Geppert1,2, W. Dreher1,2, M. Althaus1,2, D. Leibfritz1,2; 1Institut für organische Chemie, FB2, Universität Bremen, Bremen, GERMANY, 2Center for Advanced Imaging, Universität Bremen, Bremen, GERMANY.

Introduction: Routine clinical spectroscopy is mostly limited to the strong singlets of N-acetylaspartate, creatine and choline. Detection of other metabolites usually requires spectral editing. The editing method that we are concerned with is single voxel spectroscopy (SVS) double quantum filtering (DQF), based on a PRESS sequence, as described e.g. in [1]. The technique applies to coupled resonances (multiplets). A basic DQF sequence consists of a symmetric 90-180-90 pulse sandwich to create the double quantum coherence, followed by a 90 pulse to read the coherence back and a 180 to rephase the multiplet. Methods: In a PRESS-DQF implementation, the first two (90-180) and the last (180) pulses are provided by the SVS localization. The actual filtering can be performed through phase cycling, or alternatively by adding two filtering gradients to the sequence, one in the 90-90 interval and another just before data readout. It is often stated that gradient filtering is to be preferred, since it provides DQF in a single shot. This however is only true in the ideal case, and additional phase cycling is always applied, to average out unwanted signals. Here we study in some detail one possible (and on our system the most important) origin of such spurious signals. Results: We show that the four non-slice-selective 90 pulses in two consecutive shots plus the last 180 pulse form a singlet stimulated echo that survives all filtering and spoiler gradients, by a mechanism similar to that described in [2]. Furthermore, that the SVS shimming may create an important field variation inside the slice (but outside the VOI) excited by the last 180. Water singlet signals therefore appear in a band of frequencies other than 4.7 ppm and remain unaffected by water suppression pulses. Without further action, these large signals will mask the weak metabolite resonances, even after phase cycling. Conclusion: The simplest remedy is to decrease the sequence repetition rate, in order to attenuate the stimulated echo through T1-relaxation. But since the averaging of several hundred scans is usually required, this is rarely practical. As we will show, increase of the filter gradient moments provides a suitable alternative, based on echo attenuation through diffusion. So indeed gradient filtering is the preferred method of DQF, but not for the reason usually given. References: 1. Jouvensal L, Carlier PG, Bloch G, (1996) MRM 36:487-490 2. Jung KJ, Katz J, Boxt LM, Hilal SK, Cho ZH (1995) J.Magn.Reson. B107:235-241

Introduction: Recently the principle of steady state free precession (SSFP) was successfully applied to spectroscopic imaging (SI) in 31P [1] and 1H [2]. A modification of the CE-FAST sequence [3] was suggested in [2] where chemical shift selective RF pulses are used to achieve minimized repetition times (TR) while obtaining good SNR and inherently suppressing signals from water and lipids. Here the behavior of J-coupled spins in CE-FAST SI is investigated for different TR. Subjects and Methods: The sequence is depicted in Fig 1. Composite pulses as suggested by [4] are used for excitation/ refocussing using a flip angle of 40°, a complex data matrix of 16x16x16x256 (kx, ky, kz, kω) points and 10 kHz sweep width. TR was varied between 62-112 ms to investigate signals of J-coupled spins, e.g. from the metabolites myo-Inositol (Ins) or taurine. All experiments were performed on a 4.7T/40cm Bruker Biospec system with BGA-20 gradient system. A saddle type resonator was used for RF transmission and detection. Data post-processing comprised apodization, zero filling to 2048 points, 3D FFT and constant phase correction. Results: Fig. 2 displays a set of phase corrected spectra from a spherical phantom containing an Ins (75mM)/NAA (25mM) solution. Signals of Ins are observed at all TR, despite the signal modulation as seen for the {2}CH signal at 4.06 ppm in dependence on TR. All data were acquired with a constant acquisition window of 51.2 ms to provide better comparability. For the same reason a correction for unit measurement time was omitted. Discussion/Conclusion: SSFP spectroscopic imaging is a novel technique which allows to detect signals of both uncoupled and Jcoupled spins. Since there is no simulation approach of J-coupled signals concerning SSFP sequences, experimental studies will be necessary to optimize sequence parameters for coupled spins. However it was shown that signals of J-coupled signals can be observed for a large range of TR. References: 1. Speck O et al.[2002] Magn.Reson.Med. 48: 633-639 2. Dreher W et al, MRM; in press 3. Gyngell ML [1988] Magn.Reson.Imaging; 6: 415-419 4. Starcuk Z et al. [1986] J.Magn.Reson. 66: 391-397

S208

Spectroscopy: Processing and Quantification field from the z-axis (25 kHz offset, vanishing amplitude) through the x-axis (vanishing offset, 8 kHz amplitude) towards the -z-axis (-25 kHz offset, vanishing amplitude) is used. Besides improved bandwidth and tolerance to field inhomogeneity, the power consumption is almost halfened. Discussion: The novel approach enforces adiabatic dynamics by the requirement of constant adiabaticity after successively applied frame transformations. Convergence is fast and tailoring of adiabatic pulses for a wide range of applicaions is possible by additional constraints. References: 1. Garwood M, et al [2001] J. Magn. Reson. 153:155-177

395 Adiabatic following beyond the adiabatic approximation L. Mitschang, F. Schubert, H. Rinneberg; Biomedical Optics and NMR-Measuring Techniques, Physikalisch-Technische Bundesanstalt, Berlin, GERMANY. Introduction: Adiabatic passage is used to invert spins over a specified frequency range even if the acting fields are inhomogeneous. Usually, spins are forced to follow the effective field by applying modulated irradiation, obeying the adiabatic approximation in a frame of reference rotating at the instantaneous frequency of the pulse (frequency frame) [1]. This approach is shown to be far too limited. Adiabatic following can be improved by forcing the system to follow not the actual effective field but rather a related quantity. Methods: Consider a spin subjected to a rf pulse under condition of constant adiabaticity in frequency frame (subscript 0), da0/dt=k0r0, where r0 is the magnitude of the effective field, a0 its inclination to the z-axis and k0=const is the usual adiabaticity parameter. a0 and r0 depend only on the pulse amplitude and carrier frequency as functions of time. Similarly, constant adiabaticity referring to the toggling frame (subscript 1) whose z-axis points at any instant along the frequency-frame effective field implies da1/dt=k1r1. It can be shown that k0
Spectroscopy: Processing and Quantification 396 Reduction of long to short free induction decays with no loss of information D. Belkic; Medical Radiation Physics, Karolinska Institute, Stockholm, SWEDEN. Introduction: A long length N of free induction decay (FID) causes mathematical ill-conditioning in all parametric estimators from Magnetic Resonance Spectroscopy (MRS). Therefore, dimensionality reduction of the original problem is needed, but with no loss of information in the selected window. This can be accomplished by the band-limited decimation (bld)1. The fast Fourier transform (FFT) of the short bld-FID produces the same local spectrum as with the original FID. The bld-FID can be spectrally analyzed by any estimator. We use the linear predictor (LP), signal diagonalization (SD) and Padé approximant (PA) to compute magnetic resonance spectra from clinical FIDs encoded at 4T and 7T. Methods/Subjects: Given an original FID of length N and sampling time t, the band-limited decimation works as follows. The whole Nyquist interval is split into M windows. A band-limited (bl) FFT is produced in a chosen window by zeroing out the FFT outside the actual bandwidth (BW). After shifting the bl-FFT symmetrically around the zero frequency, the inverse FFT follows, yielding the bl-FID. The BW of the bl-FID is M times shorter than in the original FID. Therefore, the bl-FID can be re-sampled with M times larger sampling time than t. This yields a bld-FID of M times shorter length than N. By construction, the created bld-FID and the original FID yield the same FFT spectra in the selected window.

Spectroscopy: Processing and Quantification Results: We take two 2K FIDs with N=2048 encoded in a healthy human brain at 4T and 7T and select a window from 0 to ~5ppm. The length of the corresponding bld-FIDs is only 0.25K. Nevertheless, the original FIDs and eightfold shorter bld-FIDs are verified to yield the same FFT spectra from 0 to ~5 ppm. In Figs. 1 and 2, the LP, SD and PA are applied to these bld-FIDs yielding indistinguishable shape spectra and the identical complex frequencies/amplitudes to within machine accuracy for every retrieved metabolite. Conclusion: Mathematical ill-conditioning in quantification in MRS can be substantially mitigated by converting large to small data matrices filled with FIDs. This is achieved by the band-limited decimation which generates short FIDs with maximally 256 points that contain the whole information from 0 to ~5ppm. Such new FIDs are presently analyzed by three parametric processors. 1Dz Belkic et al. J. Chem. Phys. A 104, 11677-11684 (2000)

S209

was used as internal reference (1). Two different approaches were used to calculate metabolites concentration. Individual T2: metabolites and water T2 obtained for each subject was used to calculate metabolite concentrations for each subject. Mean T2: T2 calculated from each subject for each metabolite and water was averaged, and the mean T2 used to calculate concentrations in each subject at each TE. Results: Table reports metabolite concentrations and the relative errors calculated in the mid-brain parietal occipital cortex GM using two different methods. err% = SD/mean %. T2 (mean ± SD, ms): NAA=266 ± 34; Cr=165 ± 15; Cho = 239 ± 30; Water 125 ± 15. TE(ms)

NAA (mM) mean err% mean T2

Cr (mM) mean mean T2

err%

Cho (mM) mean err% mean T2

35

16.65

12

11.94

8

2.53

9

70

14.28

11

12.88

9

2.22

8

100

15.59

13

13.71

11

2.56

10

144

17.47

18

13.69

21

2.90

20

288

16.63

36

13.31

36

2.55

36

Individual T2

Individual T2

Individual T2

15.74

12.78

2.49

6

8

9

Similar results were obtained in the left parietal-occipital WM. Conclusion: Main results are: a) the “mean T2“ method gives different err% as a function of TE, being smaller at low TEs; b) the “mean T2“ method gives comparable results to “individual T2“ method only for Cho and Cr at TE = 35,70 ms. Our data show that low TE values are to be preferred for the calculation of the absolute metabolite concentrations when using the “mean T2“ method, in contrast to what recently shown by some authors (2). Acknowledgements: Study supported by MIUR, COFIN ex-40% 2000; Fondazione Cassa di Risparmio in Bologna; Università di Bologna, Progetto Pluriennale di Ricerca E. F. 2000. References: 1. Keevil SF, et al. [1995] Magn. Reson. Imag. 13:139-157 ; 2. Rutgers DR et al. [2002] NMR Biomed. 15:215-221.

398

397 1H-MRS cerebral metabolites quantification in human grey and white matter: analysis of different approaches for T2 relaxation correction V. Clementi1, S. Iotti2, R. Lodi2, B. Barbiroli2; 1Dipartimento di Medicina Clinica e Biotecnologia Applicata, Università di Bologna, Bologna, ITALY, 2Medicina Clinica e Biotecnologia Applicata, Università di Bologna, Bologna, ITALY. Introduction: We analysed and compared two different approaches of 1H-MRS T2 relaxation correction, used to assess cerebral metabolite concentrations in both WM and GM in 10 healthy subjects. Methods: 1H-MRS was performed on a General Electric Signa Horizon 1.5T with a standard quadrature head coil and PRESS sequence (TR/TE 4000/35, 70, 100, 144, 288 ms). Cerebral water

Quantification of Glutamate using Single Voxel Spectroscopy: Potential pitfalls A. Stengel1, T. Neumann-Haefelin1, H. Lanfermann2, F. Zanella2, U. Pilatus2; 1Department of Neurology, JW Goethe University, Frankfurt am Main, GERMANY, 2Institute of Neuroradiology, JW Goethe University, Frankfurt am Main, GERMANY. Introduction and Purpose: Short TE single voxel spectroscopy can be used to quantify metabolites in the human brain offering the opportunity for studying the pathophysiology. In contrast to long TE (>136 ms) the short TE allows quantification of nucleui with rather complex coupling patterns like Glu, Gln, mI. Due to large overlap, for the signals of Glu and Gln only the sum (Glx) is evaluated. A rather large scattering of the Glx (Glu, Gln) data (±17% ±39%) is observed [1], which may be attributed to a low SNR and baseline distorsions at short TE. However, pulse imperfections and spatial interferences, which have to be considered in long TE experiments of coupled refocussing systems like Lac [2], should also occur. While analytical description of this processes is complex, we performed a phantom study in which we investigated the impact of voxel size and shape on normalized signal intensities.

S210

Spectroscopy: Processing and Quantification

Methods: The SVS examinations (TE/TR 30/10000) were performed on a 1.5 T Philips Gyroscan Intera scanner using PRESS volume selection. The FID was digitized with 512 data points and a spectral width of 1000 Hz. Water suppression was accomplished with a CHESS pulse with a bandwidth of 60 Hz. A phantom consisting of 30 mM Glu, 30 mM mI and and 30 mM formic acid (sodium salt) was studied for quantitation purposes. The signals intensities were determined for voxels of different sizes (volume: 5,49-166,38 ml; x, y, z: 1,2-5,5 cm) and quantified in the time domain with the jMRUI software using prior knowledge. Results: Minor effects of voxel shape can be observed for formic acid (± 11%) and mI (± 10 %), while normalized Glu shows significant, reproducible variations (up to 100 %, fig) depending on voxel dimensions x, y, z. This scattering is not caused by any potential errors from the fitting procedure. Discussion and Conclusion: In vivo measured Glu data can depend on voxel size and shape. Especially when the slab excited by the first (90°) pulse is large, errors from spatial interference effects and pulse imperfections should be considered in localized spectroscopy of Glu metabolism. References: 1. G. Helms. Magn Reson Imaging 17, 1999. 2. D. Yablonskiy. MRM 39, 1998.

echo. The aim of our study was to evaluate whether reliable absolut quantification is still possible. Measurements: The TSI examinations were performed on a 1.5 T Philips Gyroscan Intera scanner using the following acquisition parameters: TR=1500ms, TE= 144, 288, 432, 576ms, FOV 240mm, slice thickness 15mm, 32 x 32 phase encoding, 2250 Hz bandwidth, 256 data points. The total measurement time was 6 minutes. Volume preselection and multiple outer volume suppression were used to suppress lipid signals from the skull. For quantitation purposes, the fraction of GM, WM and CSF in each spectroscopic voxel was determined using high-resolved Turbo IR images. Results: Peak integration was performed in the time domain with the csx-software (by courtesy of P.Barker, JHU Baltimore, MD). Absolute values for metabolite concentrations were determined using a phantom with 40 mM NAA and 40 mM lactate as reference. In vivo data were calculated by correcting for different T1 and T2 relaxation times as well as RF coil loading. The correction factors for the different relaxation times were determined experimentally in phantom studies. Coil loading correction was based on the RF power required to achieve a 90° pulse. In phantoms with different ionic strength, a good inverse linear correlation between signal intensity and coil loading was found. The quantitation method was validated in phantoms. Calibration parameters obtained from the phantom study were used to determine the NAA concentration in healthy volunteers (n=5). Tissue concentrations obtained after segmentation are shown in table1. Since lactate can be quantified by the same means, TSI is the ideal tool for assesment of metabolic changes in acute stroke. Conclusions: A modified phantom replacement method [2] for individual phase encoded multiple spin-echo experiments was presented. This methodology allows absolute quantification of NAA and Lac in 6 minutes. References: 1. JH Duyn. MRM 30, 1993 2. BJ Soher. MRM 35, 1996 NAA[mM] WM GM

frontal 8,4 +/- 0,61 8,6 +/- 0,42

parietal 9,7 +/- 0,47 8,5 +/- 0,44

occipital 8,3 +/- 0,19 9,2 +/- 0,92

400 Potentials and limitations of post-processing correction of frequency and phase shifts in in vivo magnetic resonance spectra F. Schubert, C. Elster, A. Link, M. Walzel, F. Seifert, H. Rinneberg; Medical Physics and Information Technology, Physikalisch-Technische Bundesanstalt, Berlin, GERMANY.

399 Quantitative Turbo Spectroscopic Imaging: Methodology and reliability A. Stengel1, T. Neumann-Haefelin1, H. Lanfermann2, F. Zanella2, U. Pilatus2; 1Department of Neurology, JW Goethe University, Frankfurt am Main, GERMANY, 2Institute of Neuroradiology, JW Goethe University, Frankfurt am Main, GERMANY. Introduction and Purposes: By acquiring four individual phase encoded spin-echos within a single repetition time (Turbo Spectroscopic Imaging, TSI), data acquisition time for spectroscopic imaging can be reduced, drastically [1]. Quantitative analysis of resulting spectra is hampered by ambiguous TE as well as by poor resolution caused by relatively short digitization time for each

Introduction: A signal-to-noise level sufficient for reliable quantitative analysis of MR spectra requires averaging of scans. To compensate for scan misalignment in frequency and phase due to subject motion and system instabilities [1-4], post-processing have been proposed [1-3]. We employed a potent realignment procedure [2] to study the conditions under which such methods lead to corruption rather than true improvement of the spectra. Methods: The post-processing scheme aims at optimal alignment in a selected frequency range [2]. The realignment values are determined iteratively by maximizing the modulus of the complex correlation coefficient of the shifted individual signals and their average. For analysis of the procedure, PRESS spectra were acquired from volunteer brain voxels (25x40x20 mm3) at 3 Tesla (MEDSPEC 30/100, Bruker Medical). Eight spectra of 16 phase cycled scans each were recorded (TR=3 s, TE=80 ms). Different SNRs were produced by adding real measurement noise multiplied

Spectroscopy: Processing and Quantification by 0.5, 1.0 and 3.0 (giving large, medium and small SNR). Experimental shifts were imposed by adding frequency shifts between -1 and 1, -3 and 3, -6 and 6 Hz (n=1; 3; 6, see figure) and phase shifts between -pi/8 and pi/8, -pi/4 and pi/4, and -pi/2 and pi/2. Results: The figure shows realignment results for spectra generated as described above. For each frequency shift range 3 random realizations were calculated. E.g., for spectra with medium SNR and a frequency shift between -3 Hz and 3 Hz, the error by frequency misalignment resulting from simple averaging is 1 Hz larger than that obtained by the realigment method. Values below 0 reflect a relative degradation of the final spectra by the realignment method. Thus, the study of the dependence on SNR and phase and frequency shifts reveals the conditions where attempts to realign the measured spectra start to degrade the results rather than improving them. Discussion: Our study shows that there are cases where no improvement of the spectra is gained by realignment. The risk of deteriorating the quality of the spectra is particularly high at low SNR in conjunction with mild frequency and phase shifts, i.e. in such cases it may not be advisable to apply post-processing correction. References: 1. Maudsley, AA et al, Magn Reson Imaging 1992; 10: 471-485. 2. Schubert, F et al, Proc ESMRMB 2000; 11: 189. 3. van Huffel, S et al, Proc ISMRM 2001; 9: 1682. 4. Henry, PG et al, Magn Reson Med 1999; 42: 636-642.

S211

and volume preselected PRESS-2D CSI (TE/TR/NA =135ms/1500ms/1,16x16 voxels,160x160mm FOV) sequence. Spectra from CSI voxel matching PRESS voxel were processed using CULICH software [1] with LCModel [2]. CSI data were processed twice: without apodization and using Hamming filter 0.54+0.46cos(2πl); l=-0.375 to 0.375. Correction for chemical shift artifact (CSA) was performed by adjusting the CSI grid position for each metabolite in the spectrum separately based on the fixed slice selection gradient (5mT/m) for SVS. LCModel basis set used for all spectra evaluation was acquired from SVS measurements by PRESS (TE/TR/NA=135ms/1500ms/512). Results: Spectra acquired by CSI and SVS had comparable signal/noise ratio. Error values of concentrations reported by LCModel were bellow 20%. Since the same timing of SVS and CSI sequences no additional relaxation correction was performed for the needs of comparison. The results of Student t-test (performed after F-test results)obtained for concentrations of Cholin (Cho), Creatin (Cr) and N-acetyl aspartate (NAA) in laboratory units as well as metabolite signal ratios computed by both methods are summarized in table 1. Values of metabolite concentrations in laboratory units are summarized in table 2. Table 1 T-test results for CSI and SVS comparison. No filter No filter Hamming filter Hamming filter t statistical t critical t statistical t critical (P=95%) (P=95%) Cr 4.35 2.06 2.40 2.06 Cho

0.55

2.09

1.98

2.06

NAA

1.41

2.07

1.50

2.07

Cho/Cr

4.07

2.06

1.02

2.06

NAA/Cr

2.60

2.06

0.86

2.06

NAA/Cho

1.03

2.07

0.19

2.06

Table 2.Concentrations of metabolites in laboratory units

401 LCModel for quantitative single voxel spectroscopy and spectroscopic imaging: A comparison F. Jiru, A. Skoch, M. Hajek; MR-Unit, Dept. Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague, CZECH REPUBLIC. Purpose: Both Single Voxel Spectroscopy (1H SVS) and Spectroscopic Imaging (CSI) are used for studying human brain biochemistry in vivo. Clinical application of 1H MRS requires a robust and reliable algorithms for the spectra evaluation. Of particular interest is computation of absolute concentrations of metabolites. SVS sequences can be considered as ‘gold standard' in this respect. However CSI methods are being used more frequently for quantitative analysis nowadays. In this study we compared concentrations measured both by CSI and SVS sequences and evaluated by LCModel. Subjects/Methods: All measurements were performed on 1.5T Siemens Vision scanner.13 informed healthy volunteers (21±3 years) were examined both by Single Voxel PRESS (TE/TR/NA=135ms/1500ms/256,10x10x20mm VOI) sequence

CSI with no filter CSI with Hamming filter

SVS

Cho

1.3(0.2)

1.2(0.2)

1.3(0.1)

Cr

2.5(0.5)

2.9(0.5)

3.3(0.4)

NAA

5.4(0.8)

5.4(0.8)

6(1)

Discussion: From comparison of ratios it is apparent that introducing of k-space filter led to improvement of concentration ratios agreement. However concentrations of Cr differed significantly even after filter has been applied. Cr has two visible resonances with TE=135ms (at 3.01ppm and 3.9ppm). For CSA correction dominating CH3 resonance (3.01ppm) was used. This may lead to partial volume effect since overall Cr SVS signal does not match corresponding CSI voxel. Generally concentrations computed by CSI technique are dependent also on the applied k-space filter and should be used with care. The study was supported by grant CEZ:L17/98:00023001 References: 1. Jiru et al. [2002] MAGMA; 15 suppl.1:368 2. Provencher [1993] Magn Reson Med; 30:672-9

S212

Spectroscopy: Processing and Quantification

402 Metabolite phantom correction improves 1H-MRSI diagnosis in temporal lobe epilepsy J. Lundbom1, E. Gaily2, M. Granström2, K. Vuori1, N. Lundbom1, A. Häkkinen3, S. Heikkinen1; 1Department of Radiology, Helsinki University Central Hospital, Helsinki, FINLAND, 2Department of Child Neurology, Helsinki University Central Hospital, Helsinki, FINLAND, 3Department of Oncology, Helsinki University Central Hospital, Helsinki, FINLAND. Introduction: Spectroscopic Imaging can lateralize temporal lobe epilepsy (TLE) by measuring the distribution of brain metabolites. An even metabolite distribution has been shown between normal hippocampi (HC) [1]. The PRESS-CSI sequence has a non-uniform excitation profile [2] that, together with the chemical shift, distorts the regional variation of the metabolites. The volume displacement of the metabolites varies linearly with the VOI size and has to be corrected for. We used a metabolite phantom to correct for these effects and evaluated the effect of this correction in the analysis of TLE patients. Methods: Fourteen patients evaluated with EEG, MRI, video-EEG and clinical follow-up and 8 control subjects were examined. A Siemens Vision 1.5T MR imager was used with PRESS-CSI sequence, TE/TR 270/2600, 16cm x 16cm FOV with 16x16 phase encoding steps. The VOI (15x60x80/100 mm) covered symmetrically the right (R) and left (L) HC. To correct for the imperfect excitation, a homogeneous phantom containing NAA, Cho and Cr was measured with identical parameters. The metabolite intensities of the subjects were normalized for the respective phantom values in the corresponding voxels. Results: PHANTOM: In all three metabolites, the intensity of R was higher than L. The average asymmetries for 8 and 10 cm VOI's are presented in Table 1. Table 1. The (R-L)/(R+L) difference in the phantom. VOI width Cho Cr NAA NAA/(Cho+Cr) 8cm 4.2% 4.2% 8.9% 4.7% 10cm 2.5% 3.9% 5.3% 1.9% Controls: Table 2. shows mean and SD of NAA, Cho+Cr and NAA/(Cho+Cr) before and after the correction; the mean R/L approached one and SD decreased. The maximal asymmetry limits (C-I)/(C+I) were 8% and 11 %, where C is NAA/(Cho+Cr) in the contralateral and I in the ipsilateral HC. The lower NAA/(Cho+Cr) defined I. Table 2. The change in the mean and SD of the R/L in controls with the correction. Metabolite Mean SD NAA 1.15!0.98 0.07!0.05 Cho+Cr 1.10!1.02 0.09!0.08 NAA/(Cho+Cr) 1.05!0.97 0.06!0.06 Patients: We evaluated (C-I)/(C+I) and compared it to the maximal asymmetry limit of the controls. In ten patients, the classification did not change. In three patients, the asymmetry decreased to 7%, 9 % and 5% from 9%, 14% and 10% respectively, and in one patient, increased to 12 % from 7% with the correction, thereby altering the classification. Conclusion: In evaluating TLE patients, metabolite asymmetry between the HC needs to be exactly known. Our phantom replacement method corrects the non-uniformity of the excitation profile of the PRESS-CSI sequence and allows direct R/L comparisons in the patient population.

References: 1. McLean, M. A. et al. (2001) NeuroImage 14:501-509 2. Weber-Fahr, W. et al. (2002) NeuroImage 16:49-60

403 MRS monitoring of response to 17AAG treatment: From the laboratory to the clinic M. Beloueche-Babari1, G. S. Payne1, U. Banerji2, D. J. Collins1, A. S. K. Dzik-Jurasz1, P. Workman2, I. R. Judson2, M. O. Leach1, S. M. Ronen1; 1Cancer Research UK Clinical MR Research Group, Institute of Cancer Research and Royal Marsden NHS Trust, Surrey, UNITED KINGDOM, 2Cancer Research UK Centre for Cancer Therapeutics, Institute of Cancer Research, Surrey, UNITED KINGDOM. Introduction: 17-Allylamino, 17-Demethoxygeldanamycin (17AAG) is a novel anti-tumour agent that blocks heat shock protein 90 (HSP90) causing depletion of its client proteins, including Raf-1 and Akt, and induction of HSP70. Using 31P MRS, we have shown that inhibition with 17AAG causes an increase in phosphocholine and glycerophosphocholine (GPC) levels in human colon and breast cancer cells (1, 2). The aim of this study was to test whether MRS could detect surrogate markers of the molecular and potentially clinical response to 17AAG in patients taking part in a phase I clinical trial. Methods: 1H decoupled 31P 3-D chemical shift imaging (CSI) was applied to 2 patients. This was performed on a 1.5 T Siemens Vision clinical imager/spectrometer using 8cm (for patient 1) and 12 cm (for patient 2) 31P surface coils with 1H butterfly coil, a 1s repetition time, volumes of interest of (240 mm)3 and (480 mm)3 for patient 1 and 2 respectively, and 8 phase encoding steps in 3-D giving voxel sizes of 27cc and 216cc for patient 1 and 2 respectively. Localised 31P MR spectra from the tumour volume were acquired prior to and following administration of 320 mg/m2/week of 17AAG (NCI) and processed using MRUI. Results: Patient 1 had renal cell carcinoma. Pre- and 24h post treatment scans acquired from a sternal mass showed no differences in tumour 31P metabolite content. Tumour biopsy indicated that 17AAG caused HSP70 induction while no change in Raf-1 levels was observed. Patient 2 had melanoma. Comparison of pre- and 5 days post -treatment scans from an abdominal mass showed a visible increase in the intensity of the phosphodiester (PDE) signal comprising GPC and glycerophosphoethanolamine (Fig.1). A drop in the amplitude of all the remaining signals was also detected without any significant change in the phosphomonoesters/NTP ratio. Tumour biopsy showed HSP70 induction and Raf-1 depletion following 17AAG treatment. At the time of the post-treatment scan, this patient had been on treatment for ~6 months and was considered to have clinically stable disease. Conclusions: Although our investigations need to be extended to more patients, our preliminary results indicate that, in agreement with our in vitro findings, the PDE signal could be used as a potential surrogate marker for non-invasive monitoring of response to 17AAG in patients. References: 1. M. Beloueche, et al. (2002). Proc. of ESMRMB 9,182, Cannes. 2. Y-L. Chung, et al. JNCI (submitted). Funding: Cancer Research UK (grant C1060/A808).

Animal Models

S213

growth, cells from both monolayer and spheroid cultures were examined for their growth characterisitics and found to be very similar. The results seem to indicate that, in fact, the type of spatial array determines specific changes in cell function. In particular, both energetic (e.g., alanine and lactate) as well as lipid (e.g., choline-containing metabolites and CH2 and CH3 mobile lipids) metabolism is different depending on the type of cell model utilized.

Animal Models 405

404 1H-NMR shows that monolayer cells and three-dimensional spheroids of MG-63 osteosarcoma cells have a different metabolic profile M. Santini1, G. Rainaldi1, A. Ferrante1, A. Motta2, R. Romano3, P. Indovina3; 1Laboratorio di Ultrastrutture, Istituto Superiore di Sanità, Rome, ITALY, 2Istituto di Chimica Biomolecolare, Consiglio Nazionale delle Ricerche, Naples, ITALY, 3Dipartimento di Scienze Fisiche, Università di Napoli "Federico II", Naples, ITALY. Two- and three-dimensional in vitro cultures of tumor cells have yielded much valuable information regarding various aspects of tumor biology. In particular, monolayer cultures have provided a deeper understanding principally of the underlying cellular mechanisms responsible for tumor cell behavior while three-dimensional ones (i.e., multicellular spheroids) have permitted the study of cellcell interactions in a context more closely resembling tumor tissue architecture. It is apparent that the knowledge obtained from these two cell culture systems with such different cellular organizations is complementary and that the study of both can give a more comprehensive view of tumor cell biology as a whole. High resolution nuclear magnetic resonance (NMR) spectroscopy is an extremely useful tool in examining, non-invasely, tumor cells grown both in monolayer and as three-dimensional spheroids. For example, this technique has been applied to examining tumor cell growth, membrane lipid metabolism, efficacy of various treatments and cholesterol metabolism in monolayer cultures as well as metabolism and response to anti-neoplastic agents in three-dimensional tumor spheroids. As can be discerned from a careful analysis of these and other papers which use NMR, much important information regarding cell function and cellular organization can be obtained by this technique. However, as is also apparent, little attempt has been made to compare and integrate the complementary information which emerges from monolayer and spheroid cultures. It was the purpose of the present paper to attempt to compare monolayer and spheroid cultures of the same cell type (MG-63 human osteosarcoma cells) in order to determine if the spatial organization of cells can induce important variations. For this task, the high resolution proton NMR (1H-NMR) spectra were obtained from both types of cultures and then compared by utilizing a new algorithm developed by our group. In order to make certain that the differences observed were not due to the effects of hypoxia, small spheroids (about 50 - 80 µm in diameter) with no hypoxic center were used. In addition, in order to exclude the effects of cell

ρ spectroscopy of rat brain metabolites in vivo T1ρ M. I. Kettunen1, O. H. J. Gröhn1, R. A. Kauppinen2; 1Department of Biomedical NMR, University of Kuopio, Kuopio, FINLAND, 2School of Biological Sciences, University of Manchester, Manchester, UNITED KINGDOM. Introduction: Recent data indicate that T1ρ of water offers novel information on tissue status in experimental cerebral ischaemia (12). Alterations in proton exchange, cross-relaxation and microsusceptibility are likely to contribute to observed T1ρ response. In this study, T1ρ relaxation times of brain metabolites, namely N-acetylaspartate (NAA), creatine (Cr) and lactate (Lac), were measured in normal rat brain and following cardiac arrest. Methods: Halothane-anaesthetised male Wistar rats (n=6) were studied in a 4.7T magnet (Magnex) interfaced to a Varian UNITYINOVA console, using a quadrature surface coil. A 3*10*6mm3 spectroscopic voxel was placed to fronto-parietal region and T1ρ was quantified using an adiabatic PRESS-type localisation (3) pulse sequence where the excitation pulse was replaced with spinlock pulse consisting of 4 ms adiabatic half passage and a cw pulse. Water suppression was performed using Watergate after localisation part. The frequency of the spin-lock pulse was adjusted to either NAA, Cr or Lac frequency. Six spin-lock times (10-510ms) and two spin-lock amplitudes (0.5G and 1.0G) were used with TR 4s, TE 66ms and 32 averages. Additionally, T1ρ of water was measured using five spin-lock times (10-90ms) with TR 4s, TE 56ms and 2 averages. After acquisition of control spectra, cardiac arrest was induced using KCl and 10 minutes later, post-mortem spectra (n=4) were collected in the order: H2O, NAA, Cr, Lac. Results: All the metabolites showed apparently monoexpontential signal decay within the spin-lock time range studied. Table 1 shows the measured relaxation times for the metabolites. Following cardiac arrest, water T1ρ increased 5-8%, while NAA T1ρ decreased by 30% and Cr T1ρ did not change. Discussion: Consistent with previous water T1ρ results (1-2), the obtained in vivo metabolite T1ρ values are longer than values previously reported for T2 or Carr-Purcell-T2 at 4T (3). Interestingly, T1ρ response to ischaemia differs between water and metabolites, suggesting that these molecules sense different physico-chemical mechanisms, including proton exchange and cross-relaxation, in the post mortem brain. Known differences in magnitudes of MT observed in brain (4) and compartmentation of metabolites may partly explain the varied response of metabolite T1ρ. References: 1. Gröhn OHJ et al. [1999] Magn.Reson.Med 42:268-276. 2. Kettunen MI et al.[2001] Magn.Reson.Med 46:565-572. 3. Michaeli S et al. [2002] Magn.Reson.Med 47:629-633 . 4. DeGraaf RA et al. [1999] Magn.Reson.Med 41:1136-1144.

Animal Models

S214 T1ρ relaxation times in normal and in post mortem brain

Control (0.5G) Post Mortem(0.5G) Control (1.0G) Post Mortem (1.0G) Water 79+/-1

83+/-1*

90+/-1

98+/-1*

NAA 570+/-29

391+/-54*

607+/-35

414+/-27*

Cr

437+/-25

441+/-48

421+/-31

465+/-51

Lac

n.d.

325+/-23

n.d.

341+/-64

References: 1. Wu O, et al [2001] Stroke 32:933-942. 2. Dijkhuizen RM, et al. [2003] Proc of Brain 03 (abstr). 3. Dijkhuizen RM, et al [2002] JCBFM 21:964-971.

406 Predicting histological outcome in a rat acute embolic stroke model using combined DWI and PWI O. Wu1,2,3, T. Sumii4, M. Asahi4, M. Sasamata5, A. Sorensen3, L. Ostergaard2, B. R. Rosen3, E. H. Lo4, R. M. Dijkhuizen1,3,4; 1Image Sciences Institute, University Medical Center Utrecht, Utrecht, NETHERLANDS, 2CFIN Dept of Neuroradiology, Arhus University, Arhus, DENMARK, 3A A Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, 4Neuroprotection Research Laboratory, Massachusetts General Hospital, Charlestown, MA, 5Institute for Drug Discovery Research, Yamanouchi Pharmaceutical Co., Ltd., Tsukuba, Ibakari, JAPAN. Purpose/Introduction: Tissue signature maps using acute diffusion-weighted (DWI) and perfusion-weighted (PWI) MRI have been shown previously to predict tissue outcome in acute human stroke [1]. However, a limitation of that study was the reliance on follow-up imaging to provide an indirect measurement of tissue infarction. The purpose of this study is to extend predictive algorithms to experimental animal models of stroke where extent of infarction can be validated histologically. Subjects/Methods: Data from the control arm of a trial on thrombolytic treatment with pamiteplase, a genetically modified tissue plasminogen activator (t-PA) in Sprague-Dawley rats (n=8) after unilateral embolic stroke [2] were used to train a tissue risk model using a generalized linear model (GLM) [1] (Group 1). Derived coefficients were applied to MR data acquired from the control arm of a delayed recombinant t-PA treatment study in spontaneously hypertensive rats (n=6) [3] (Group 2). For Group 1, the training set, animals were imaged 1 h post-stroke. For Group 2, the histology comparison group, animals were imaged 4 h post-stroke. DWI and PWI were performed on a 2T NMR spectrometer [3]. Maps of apparent diffusion coefficient (ADC), cerebral blood flow (CBF), cerebral blood volume (CBV) and mean transit time (MTT) were calculated and used as GLM input. In Group 2, acute MRI and GLM risk maps were coregistered to 2,3,5-triphenyletetrazolium chloride (TTC) stained brain sections (10 h post-stroke). Sensitivity and specificity of the GLM in predicting infarction were calculated along with receiver operating characteristic (ROC) curves. Area under the ROC curve (AUC) was calculated by fivepoint Newton-Cotes numerical integration. Results: The figure shows a map of GLM predicted risk of infarction overlaid on as acute ADC image along with corresponding histology. From Group 1, the optimal operating point was found to be 23% and used to evaluate predicted infarct volumes in Group 2. Sensitivity was 94%, specificity was 85% with an AUC=.94. Discussion/Conclusion: Tissue risk maps based on acute DWI and PWI can be used to assess risk of infarction as identified on TTCstained brain sections. Since risk maps can identify risk of tissue infarction without treatment, these maps may be useful for evaluating efficacy of intervention in experimental stroke models and thereby provide a metric that could translate readily into clinical settings.

407 Early brain edema after experimental trauma detected by diffusion weighted MRI H. Lahrech, O. Carle, R. Farion, H. Reutenauer, J. F. Payen; Inserm u594, Hopital Michallon, Grenoble, FRANCE. Introduction: Elevated intracranial pressure subsequent to brain edema is the most frequent cause of death in head-injured patients. However, early imaging of brain edema following trauma has received little attention. We used diffusion-weighted MRI to determine apparent diffusion coefficients (ADCs) within 7 hours after experimental trauma, using the impact- acceleration model. This model results in diffuse brain injury (J Neurosurg, 1994;80:291). Methods: Twelve anesthetized, mechanically ventilated rats (330430 g) were studied using 7T magnet. All animals were stable prior MRI study: PaCO2 30-40 mmHg, PaO2 >140 mmHg, mean arterial blood pressure (MABP) >75 mmHg, normothermia. Diffusionweighted MRI (TE 80ms, TR 2s, Tdiff 25ms, coefficient b 500s/mm2, 5 coronal slices with a thickness of 1.5 mm each) was done in 2 groups of rats : trauma (n=6) and sham-operated (n=6). ADCs were determined at H1, H4, and H7 in intracrebral region (neocortex, corpus callosum, striatum) and in extracerebral region. Data (mean±SD) were compared using ANOVA and Student-t test. Results: No difference was found between the two groups regarding blood gas, arterial pH, hemoglobin content. MABP was lower in the trauma group (100±23 vs 120±19 mmHg. p<0.05). Extracerebral ADC remained unchanged throughout the experiment in the two groups. Intracerebral ADC was significantly lower in trauma group at H1, and subsequently decreased at H7 (figure). Discussion: We used a model of trauma brain alone, with no additional injury (e.g. arterial hypotension, hypoxia) known to sensitize ADCs changes (J Neurosurg 1996;84:97) . These conditions resulted in early brain edema, detectable at H1 post-trauma. The decrease in intracrebral ADC suggests that cellular (or cytotoxic) edema would predominate in this type of injury. Evaluation of neuroprotective procedures is therefore possible using this approach. Figure. (* p<0.05 vs sham. Ý p<0,05 vs H1)

Animal Models

S215

quence can distinguish between edematous tissue and CSF. Discussion and Conclusion: NG2 positive U251N glioblastomas vascularized and grew faster compared with wild-type tumors. The working hypothesis is that specific up-regulation of angiogenesis occur in part by NG2 sequestering of angiostatin (1). This study demonstrates the versatility of in vivo MRI in longitudinal phenotypic imaging of an orthotopic experimental tumor model. Reference: 1. Chekenya, M. et al. FASEB J 16(6): 586-8 (2002).

409 NMR characterization of apoptosis-determined changes in the hydration parameters of rat liver mitochondria D. L. M. Pouliquen1, C. Moreau1, G. Bellot1, G. Guihard1, J. P. Benoit2, K. Meflah1, F. M. Vallette1; 1Inserm UMR 419, Institut de Biologie, Nantes, FRANCE, 2Inserm ERM 0104, IBT, Angers, FRANCE.

408 NG2 receptor status modifies human glioblastoma development: assessment by MRI C. Brekken1, C. Brekke2, M. Chekenya3, P. O. Enger3, T. B. Pedersen1, A. Lundervold2, R. Bjerkvig3, O. Haraldseth1; 1Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, NORWAY, 2Physiology, University of Bergen, Bergen, NORWAY, 3Anatomy and Cell Biology, University of Bergen, Bergen, NORWAY. Purpose/Introduction: The neural stem cell receptor NG2 promotes tumor angiogenesis in vivo. In aggressive brain tumors, the NG2 core protein is co-localized with sprouting microvessels (1). In this study, MRI was performed twice during growth of NG2 positive and wild type human glioblastoma phenotypes in the nude rat brain. Subjects and Methods: The human glioblastoma cell line U251N was transfected with the rat NG2 gene as described earlier (1). Suspensions of spheroids containing ~40000 NG2 receptor positive (NG2+) or wild type (WT) cells were stereotactically injected into the right subcortex of 17 nude rats (1). 5 and 7 weeks after cell implantation, the rats underwent MRI examination. MRI was performed at 2.35 T using a Bruker Biospec Avance DBX-100 (Bruker, Germany). A polyethylene catheter was inserted into the right femoral vein for intravenous (i.v.) injection of contrast agent. Total experimental time was ~130 minutes. Nine different 2D MR pulse sequences were employed. 0.2 mmol/kg of Gadomer-17 (kindly provided by Dr. Misselwitz, Schering AG, Berlin, Germany) was injected as a short bolus after 4 initial baseline scans in a dynamic T1-weighted sequence. An additional post-contrast acquisition started 2 min after an additional bolus injection of 0.5 mmol/kg OmniscanTM (Amersham Health, Norway). Results: NG2 over-expression accelerated the development and growth of the U251N tumors in rat brain. Tumor volumes (Mean±SEM) after 5 weeks were 2±1 for WT vs. 11±5 mm3 for NG2+, and after 7 weeks, 11±6 for WT vs. 121±42 mm3 for NG2+. Animals bearing NG2+ tumors were all sacrificed within 65 days after cell implantation, whereas the majority (5/9) of the rats bearing WT tumors was alive at the day of maximal follow up time (82 days). OmniscanTM and Gadomer-17 showed different enhancement patterns, indicating distinctive roles in estimation of tumor microvascular function. T2-weighted and FLAIR, together with post OmniscanTM T1-weighted images allow estimation of volume of tumor and tumor associated edema, since the FLAIR se-

Purpose/Introduction: In the process of apoptosis, the fight for the cell's fate takes place in mitochondria, which sequester a potent cocktail of pro-apoptotic proteins (1). We have previously demonstrated that temperature-dependent analysis of proton relaxation times allows to identify changes in the hydration parameters related to functional modifications of mitochondria (2). The purpose of this study was to characterize the changes in the physical properties of mitochondrial water in the course of apoptosis. Subjects and Methods: BD IX female rats of 12 weeks of age, fasted overnight, were used for the study. Apoptosis was induced by an i.p. injection of lipopolysaccharide and galactosamine. Mitochondria were isolated from the liver of 3 controls (C) and 4 apoptotic (A) rats by an original procedure employing discontinuous Percoll gradients which allowed to separate the different populations of mitochondria according to their density. The pellets of the dense mitochondrial fractions (MH) were immediately frozen in liquid nitrogen, followed by NMR analysis on a temperatureregulated NMS 120 minispec Bruker spectrometer (20 MHz). The temperature dependency of the spin-lattice relaxation times of nonfreezing water (structured water, T1sw) was analysed from 255 K to 240 K, and then immediately after thawing the spin-lattice relaxation times and spin-spin relaxation times of total water (T1obs and T2obs) were measured at 277 K. Results: Mitochondria MH isolated from the liver of apoptotic rats exhibited a significant rise in the T1obs (364.9 / 249.1 msec) and T2obs (71.9 / 45.5 msec). Below the freezing point of bulk water, the temperature dependency of T1sw in (A) differed from (C) by 2 main features: a shift of +4 K in the temperature Tm corresponding to the minimum of T1sw, and a more pronounced increase in T1sw below Tm. Discussion/Conclusion: The rise in the relaxation times of total water means a 21 % and 26 % increase of the distance over which T1obs and T2obs are averaged, a feature in good agreement with the idea that apoptosis produces an increase of the permeability to water of the mitochondrial membranes. The changes observed in the thermodynamic properties of structured water suggest that apoptosis also produces an increase in the activation energy for the rotational motion of structured water molecules inside the hydration shell of proteins together with a lengthening of their mean rotational correlation times. References: 1. Hengartner M. O. Nature (2000) 407, 770. 2. Pouliquen D., et al. MAGMA (2002) 15 (Suppl.1), 289.

S216

Animal Models

410 In vivo MR imaging of spinal cord injury at 17.6 Tesla T. Weber1,2, T. Neuberger1, N. Weidner2, M. Vroemen2, V. Behr1, I. Wieland1, U. Bogdahn2, A. Haase1, C. Faber1; 1Institute of Physics (EP5), University of Würzburg, Würzburg, GERMANY, 2Department of Neurology, University of Regensburg, Regensburg, GERMANY. Introduction: 1H magnetic resonance imaging (MRI) is a common method to detect spinal cord pathologies in humans. In contrast, in animal models spinal cord data is usually obtained by histopathological analysis. The main reasons for this are that spinal cord MRI is hampered by limits in spatial resolutions and mostly accompanied by artefacts because of respiratory motion and blood flow. The 750 MHz wide bore magnets could overcome the first problem by providing a better signal-to-noise ratio (SNR) than lower field systems. Objects and Methods: MR imaging Imaging experiments were conducted on a Bruker AVANCE 750WB system at 17.6 Tesla. The maximum gradient strength was 200 mT/m at a rise time of 180 µs. We used a home-built linear surface coil for transmission and signal detection. To avoid artifacts caused by blood flow and respiratory motion ECG triggering was combined with breath gating. As imaging sequence a multi-slice 2D gradient echo was used with an echo time of 4.1 ms and a repetition time around 200 ms depending on the heart rate. After each heart beat one k-space line was detected for each slice, i.e., every slice was acquired at a different heart phase. No signal was detected during breathing periods. Animal preparation Adult female Fischer 344 rats received a midthoracic laminectomy followed by a contusion of the spinal cord using an IH Impactor (200 kdyn applied force). Two weeks after contusion the rats were anesthetized with 2% isofluorane with Carbogen (95 % oxygen and 5 % carbon dioxide) as respiratory gas. Results: The spinal cord images in the lower thoracic part of the spinal cord could be acquired virtually without motion artefacts at a spatial resolution of 102x102 µm² in-plane and a slice thickness of 350µm in a total scan time of 25 minutes. Spinal cord gray and white matter could be clearly distinguished in healthy tissue. At the contusion, the contrast between gray and white matter disappeared and cystic lesion defects were clearly identifiable indicating the lesion center (see also Figure 1). Discussion: The study showed that it is possible to directly visualize spinal cord injuries at 17.6 Tesla at a higher spatial resolution as compared to studies performed at lower field strengths. This will improve the quality of time course studies of spinal cord injury models since the lesions can be visualized in higher detail.

411 Energetics and histopathology assessed by 31P MRS and histochemical techniques in the early onset of rejection in concordant hamster xenohearts O. Rakotonirainy1, D. J. Lukes2, Å. Tivesten3, A. Lundgren2, C. Kjellström4, J. Isgaard3, A. Karlsson-Parra5, M. Olausson2, B. Soussi1; 1Bioenergetics Group, Wallenberg Laboratory, Sahlgrenska University Hospital, Göteborg, SWEDEN, 2Department of Transplantation and Liver Surgery, Sahlgrenska University Hospital, Göteborg, SWEDEN, 3Research Center for Endocrinology and Metabolism, Sahlgrenska University Hospital, Göteborg, SWEDEN, 4Department of Pathology, Sahlgrenska University Hospital, Göteborg, SWEDEN, 5Department of Clinical Immunology, Sahlgrenska University Hospital, Göteborg, SWEDEN. Purpose: The impact of ischemia/reperfusion and the importance of apoptosis during rejection are largely unknown in xenotransplants. The aim of the study is to investigate by high resolution 31P MRS the status of the high energy phosphates in the initiation of rejection and to determine by histochemical techniques the nature of the physical damage at this period of time. Subjects and Methods: Cervical transplantations between inbred Lewis (RT11) and Golden-Syrian hamsters were performed (n=5) and between hamsters (n=5, controls). After 30 minutes of cold ischemia and 6 hours of reperfusion the heart were explanted and divided in halves. The upper part was extracted with perchloric acid1 to be analyzed by in vitro 31P MRS. The lower part were devoted to the histological studies: Billigham's score, TUNEL2 and DNA laddering3. The 31P NMR measurements were carried out at 2770K on a Bruker Avance DMX 500MHz operating at 202.4MHz..The absolute concentrations of the metabolites were determined by comparison with an internal standard. Results: There is a significant difference in the PCr content (Figure 1), 14.80 ± 1.04 (xenoset) vs. 8.00 ± 1.18 (control) µmol/g dry weight heart tissue (mean ± SEM; p=0.02) leading to higher PCr/ATP ratios in the xenogroup. Representative spectra from each group are shown in Figure 2. The histological studies depicted necrosis and no antibodies, complement or other differences was detected.

Contrast Agents and Mechanisms

S217

Introduction: The water residence time (τM) is a critical, but not well controlled, parameter of the relaxivity of MRI contrast agents. In this study, we have investigated the influence of the charge through a series of similar complexes carrying different charges (figure 1).

Figure 1: PCr concentrations of the hamster allogeneic hearts (H) and the xenosets (X)

Figure 1: Gd-DTPA(-2) 1(R1=R2=R3=O-), Gd-DTPA-BME(0) 2(R1=O-, R2=R3=OCH3), Gd-DTPA-BMA(0) 3(R1=O-, R2=R3=NHCH3), Gd-DTPA-MMA(-1) 4(R1=R3=O-, R2=NHCH3), Gd-DTPA-PentaMethylA(+3) 5(R1=R2=R3=NHCH3), Gd-DTPA-PentaA(+3) 6(R1=R2=R3= NH2) Subjects and Methods: The bisamide and bisester complexes were obtained from DTPA-bisanhydride and the corresponding amide or alcohol.(1) The monoamide was prepared as described by Sieving et al.(2) The pentamides were obtained from the actived pentaesters (prepared in situ by successive action of the 4-dimethylaminopyridine (DMAP), the dicyclohexylcarbodiimide (DCC) and the N-hydroxysuccinimide (NHS)) and the amine in excess. Variable temperature 17O NMR measurements were recorded as already described on a Bruker AMX-300 spectrometer (7.05T).(3) Results: The water residence times of the positively charged pentamide complexes are noticeably long (table 1). Neutral bisester has an exchange rate very close to the one of the doubly negatively charged Gd-DTPA but much faster than the neutral Gd-DTPABMA (table 1). Table 1

Figure 2: Representative spectra of control (A) and xenoheart (B) showing higher PCr Conclusion: The early signs of rejection at this time point displayed features of early, possibly mitochondria necrosis.The ischemia-reperfusion insult per se did not cause any apoptosis. As evidenced from 31P MRS results, the PCr concentrations of the xenohearts were higher, perhaps due to mitochondrial disruption. These results shed new light on the rejection process in this model and warrant further investigations. References: 1. Lagerwall K, Madhu B, Daneryd P, Scherstén and Soussi B (1997) Am J Physiol., 272, 41., H83 2. Saraste A (1999) Herz, 24, 3., 189 3. Maulik N, Yoshida T, and Das D.K (1998) Free Radic Biol Med., 24, 5., 869

Contrast Agents and Mechanisms 412 Towards an improvement of the relaxivity of paramagnetic contrast agents: A study of the charge-water exchange relationship in a series of Gd-DTPA derivatives S. Laurent, F. Botteman, L. Vander Elst, R. N. Muller; Dpt Organic Chemistry/NMR Lab, University of Mons-Hainaut, Mons, BELGIUM.

Gd-DTPA 1 Gd-DTPA-BME 2 Gd-DTPA-BMA 3 Gd-DTPA-MMA 4 Gd-DTPA-PentaMethylA 5 Gd-DTPA-PentaA 6

Charge -2 0 0 -1 +3 +3

τM310 (ns) 143 150 967 178 2882 2521

Discussion and Conclusions: The similarity of water residence time of the bisester complex 2 and Gd-DTPA 1 suggests that the charge is not the factor dominating the water exchange. As shown in a previous study of bisamide derivatives(3), the presence of a hydrogen atom on the amide group allows bonds unfavorable to the exchange, a situation which does not exist for the bisester which consequently presents a more rapid exchange rate. Lowering the negative charge by replacement of carboxylate groups by amides (Gd-DTPA-MMA 4, Gd-DTPA-BMA 3, GdDTPA-PentaMethylA 5 and Gd-DTPA-PentaA 6) progressively modifies the congestion in the coordination sphere and induces a lengthening of the water residence time. The exchange rate seems therefore to be more sensitive to the sterical hindrance inside the first coordination sphere than to the charge of the complex. References: 1. Laurent S, Vander Elst L, Houzé S, Guérit N, Muller RN [2000]Helv.Chim.Acta 83:394-406 2. Sieving PF, Watson AD, Rocklage M. [1990] Bioconjugate Chem. 1:65-71 3. Botteman F, Nicolle G, Vander Elst L, Laurent S, Merbach AE, Muller RN, [2002]Eur. J. Inorg. Chem. 2686-2693

S218

Contrast Agents and Mechanisms

413

414

Synthesis and physicochemical characterization of new GdTTDA-bisamide complexes, potential contrast agents for MRI S. Laurent, L. Vander Elst, R. N. Muller; Dpt Organic Chemistry/NMR Lab, University of Mons-Hainaut, Mons, BELGIUM.

Multi-frequency EPR and NMRD characterization of small particulate gadolinium oxide and gadolinium oxide albumin microspheres M. A. McDonald1, K. L. Watkin1, R. B. Clarkson2, A. Roch3, R. N. Muller3; 1Medical Imaging Research Laboratory, Department of Speech and Hearing Science, University of Illinois, Urbana, IL, 2Department of Veterinary Clinical Medicine and the Illinois EPR Research Center (IERC), University of Illinois, Urbana, IL, 3NMR Laboratory, Department of Organic Chemistry, University of Mons-Hainaut (UMH), Mons, BELGIUM.

Introduction: Ideally at 310 K, the water residence time τM of a Gd-complex should be equal to ca 10 to 30 ns, depending on the imaging field, allowing so for the highest relaxivity increase after covalent or non-covalent binding. Recent works have shown that derivatives of Gd-TTDA have τM significantly shorter than GdDTPA (τM <10 ns) [1, 2]. This work reports the synthesis and the physicochemical characterization of 2 new Gd-TTDA-bisamide complexes which could be further substituted to enhance their specificity.

Subjects and Methods: The TTDA ligand (3,6,10-tri(carboxymethyl)-3,6,10-triazadodecanedioic acid) was prepared as described [3]. The TTDA-bisamides were obtained by reaction of TTDA bisanhydride with the corresponding amine. The Gd-complexes were prepared by mixing ligand with GdCl3. Proton longitudinal nuclear relaxation dispersion profiles (NMRD) were recorded at 310 K on a field cycling relaxometer (Stelar). τM of the complexes was obtained from the 17O transverse relaxation rate (Bruker AMX-300) [4]. Transmetallation by zinc ions was evaluated as previously described [5]. Results: At low temperature, no limitation of r1 by τM is observed in good agreement with the low τM values obtained by 17O NMR at 310 K: 22.0 ns and 32.4 ns respectively for 2 and 3. At 310 K, NMRD profiles are quite similar for the 3 complexes; r1 at 20 MHz is equal to 3.9, 4.2 and 4.0 s-1mM-1 for 1, 2 and 3 respectively. Their stability assessed by the transmetallation by Zn(II) in phosphate buffer is similar and low as compared to Gd-DTPA. Discussion and conclusions: As already observed, amide groups prolong the τM of Gd-complexes. Nevertheless, τM of both bisamide derivatives of Gd-TTDA is optimal for MRI. As such, these complexes are unfortunately rather sensitive to transmetallation. Some efforts should therefore be devoted to improve the structure of the backbone in order to improve the stability of the complexes. References: 1. Cheng T-H, Wang Y-M, Lin K-T, Liu G-C [2001] J. Chem. Soc. Dalton Trans. 3357-3366 2. Cheng T-H, Lee T-M, Ou M-H, Li C-R, Liu G-C, Wang Y-M [2002] Helv. Chim. Acta, 85:1033-1050 3. Wang Y-M, Lee C-H, Liu G-C, Sheu R-S [1998] J. Chem. Soc., Dalton Trans. 4113-4118 4. Vander Elst L, Maton F, Laurent S, et al. [1997] Magn. Reson. Med. 38:604-614 5. Laurent S, Vander Elst L, Copoix F, Muller RN [2001] Invest. Radiol. 36:115-122

Introduction/Purpose: The objective of this study was to characterize the magnetic and electronic relaxation properties of small particulate Gd2O3 (20-40 nm particles, SPGO) and SPGO embedded within albumin microspheres (GOAM): a prototype multimodality agent for MR, US, CT and gadolinium neutron capture therapy. Methods: Albumin microspheres and GOAM were synthesized using a modified water-in-oil polymerization method. SPGO concentration was determined by ICP-MS. The 1/T1 NMRD profiles of the solvent protons were obtained on a Stelar field cycling relaxometer (proton Larmor frequency range of 0.02 MHz to 10 MHz). A theoretical model of relaxivity was also utilized. Relaxometry measurements were made at 20 MHz, 40 MHz, 60 MHz, and 300 MHz using Bruker Multispec instruments at 37ºC in 0.4% agar gel. EPR studies were done using various Bruker and IERC-developed instruments. Results: A plateau was observed for the relaxivity of both SPGO and GOAM as concentration and magnetic field strength increases. This non-linear behavior may be due to the insolubility of SPGO in aqueous solution and the formation of aggregates of GOAM. No intrinsic water proton relaxation rate enhancement was observed with albumin microspheres lacking SPGO. Experimental disruption of GOAM resulted in relaxivity greater than that of SPGO but less than that of intact GOAM, indicating that un-sequestered SPGO particles interact de novo and/or remain associated with the disrupted albumin shell. The NMRD profile of GOAM revealed an increase in relaxivity and a shift of maximal proton relaxation enhancement into the higher frequency range in comparison to the NMRD profile of SPGO. GOAM showed maximal relaxivity between the 40-100 MHz frequency range. Multi-frequency EPR studies were conducted in order to determine the influence of electronic relaxation parameters (i.e., the zero field splitting interaction and its correlation time) on SPGO and GOAM relaxivity and their NMRD profiles. Discussion/Conclusions: Single frequency 1/T1 and 1/T2 proton relaxation rate measurement and NMRD analysis demonstrates the enhanced relaxivity of SPGO embedded within albumin microspheres (GOAM) when compared to albumin microspheres and SPGO. While albumin microspheres show no significant MR enhancement, nonspecific sequestration of SPGO within macromolecular albumin microspheres allows increased interaction with mobile protons and/or decreases the tumbling rate of the gadolinium nuclei further enhancing SPGOs effectiveness as an MR contrast agent. Results will be discussed in light of current multi-frequency EPR and NMRD theory. Acknowledgement: Supported by the M.J. Neer Research Fund; the IERC, UIUC, and NMR Laboratory, UMH, for EPR and NMRD studies, respectively.

Contrast Agents and Mechanisms

S219

415

416

Intratumoral Accumulation of Folate Receptor Targeted USPIO D. Rouze1, S. Laurent1, S. Bohic2, L. Vander Elst1, R. N. Muller1; 1Dpt Organic Chemistry/NMR Lab, University of Mons-Hainaut, Mons, BELGIUM, 2Micro-fluorescence - ID22, ESRF, Grenoble, FRANCE.

Water diffusion influences Magnetization Transfer Processes in a pH dependent manner M. Benito1, P. López1, A. Sierra1, T. B. Rodrigues1, P. Ballesteros2, S. Cerdan1; 1NMR Laboratory, Instituto Investigaciones Biomédicas, Madrid, SPAIN, 2Laboratorio de Síntesis Orgánica e Imagen Molecular, Instituto de Investigación, Madrid, SPAIN.

Purpose: The folate receptor (FR) is a membrane protein overexpressed in many human tumors like ovary, uterus, kidney and brain tumors. We have grafted folate molecules on USPIO with the aim of a specific internalisation of the particles via receptor-mediated endocytosis [1]. Folate molecules is a good vectorizing moiety because of its high affinity for the FR (Kd ~10-10 M), its low immunogenicity, ease of modification and small size (Mr 441.4) [2,3]. Subjects and Methods: HeLa, CHO-K1 (Chinese Hamster Ovary), and FRGPI-16 cell lines were been used. HeLa cells naturally overexpress FR while FRGPI-16 are CHO-K1 transfected with the FR. Internalisation of folate receptor-targeted nanoparticles was studied by confocal microscopy and quantified by ICPAES. Cells were grown on RPMI 1640 medium (without folic acid) containing targeted USPIO labelled with FITC (fluorescein isothiocianate). After 3 and 48 hours of incubation, cells were analysed by confocal microscopy and ICP-AES. Controls were cells incubated with non-specific USPIO. Results: The specific internalisation of folate grafted USPIO was first demonstrated by confocal microscopy after incubations of the cells during 3 and 48 hours. Fluorescence has been detected into the cytoplasm of cells overexpressing FR (HeLa and FRGPI-16) while it is weak in CHO-K1 (not expressing FR). The same experiment has been reproduced but with 1 mM folic acid into the medium. In this case, fluorescence remained outside the cells. This observation tends to demonstrate the specific internalisation of the vectorized particles by FR. Non-specific USPIO show a very weak internalisation. These observations have been confirmed by singlecell mapping with hard x-ray generated by a synchrotron (Table 1) and by ICP analysis (Figure 1) which show increased internalisation of specific particles as compared to non-targeted ones.

Conclusion: We have realised a specific vectorisation of targetedUSPIO to tumoral cells by the folate receptor pathway. Such internalisation should be sufficient to produce a significant contrast enhancement of the NMR images. 1. Antony AC [1992] Blood 79:2807-2820 2. Gruner BA, Weitman SD [1998] Invest. New Drugs 16:205219 3. Konda SD, Aref M, Wang S, Brechbiel M, Wiener EC [2001] MAGMA 12:104-13

Introduction: Magnetization Transfer (MT) between water and the exchangeable hydrogens of biomolecules (BM), constitutes a powerful source of endogenous contrast in the images obtained by Magnetic Resonance methods. Basically, two mechanisms have been proposed for MT processes. Direct exchange of magnetization between the water and the exchangeable protons of BM (OH, NH, etc…) or exchange of magnetization between water molecules with restricted rotational dynamics solvating BM and the freely moving water molecules in bulk solvent. Both mechanisms involve water diffusion but the influence of this parameter on MT has not been evaluated previously. Methods: We investigated the influence of the apparent diffusion coefficient (ADC) of water on the MT transfer of solutions of 1M L-Glutathione containing or not 50% Glycerol. Z-spectra representing the ratio of intensities Ms/Mo of the water resonance were obtained before (Mo) and after (Ms) presaturation (5s) over a range of frequencies at pH 4.0, pH 7 and pH 9.0 using a Bruker AM-360 NMR spectrometer. ADC´s for water were determined using the PFG method. Results: Z spectra depicted two overlapping MT peaks at 4.7 ppm and 8 ppm, respectively. The first MT peak corresponds to MT between water protons while the second peak reflects MT between NH groups and water. The presence of glycerol, decreased the diffusion coefficient of water from 1.5 to 0.3 10-5 cm2 s-1. These resulted in appreciable effects on the z-spectra obtained at the different pH´s. Conclusion: Recently, a variety of Magnetic Resonance Imaging methods have been proposed to measure extracellular pH using MT methods (Ward and Balaban, Mag. Res. Med 2000, 44, 799-802; Zhang et al, 1999, Angew. Chem, 38, 3192-3194). Our results indicate that, in addition to tissue pH, the diffusion coefficient of water may influence considerably the MT processes affecting inevitably the pH measurement. Therefore, in order to implement successfully these methods, appropriate correction factors must be introduced to account for the effects of the different diffusion environments. Acknowledgement: This work was financed in part with a strategic group grant from the Community of Madrid to PB.

417 Evaluation of magnetization transfer effects at 1.5T and 3T: a comparison between conventional and hyperecho-TSE's M. Weigel, J. Hennig; Section of Medical Physics, Department of Diagnostic Radiology, University Hospital of Freiburg, Freiburg, GERMANY. Purpose: Clinical routine MRI often relies on 2D turbo-spinechosequences (TSE) with a large number of slices for good volume

S220

Contrast Agents and Mechanisms

coverage. One undesirable side-effect of this acquisition is that adjacent slices act like an off-resonance excitation pulse. This leads to an “artificial MT-effect” and, therefore, undesired MT-contrast. Recently, novel techniques of flip angle variations such as hyperechoes[1] and TRAPS[2] were introduced for solving SAR-problems. Since the MT-effect is strongly correlated to the extent of flip angles used, it may be expected that such implementations of “hyperTSE-sequences” generally show a lesser MT-effect than TSE180°-sequences. This investigation presents quantitative studies for a first comparison of the MT-effect on conventional and hyperTSE-sequences. Subjects and Methods: All experiments were performed on two whole-body-imaging systems (1.5T/3.0T, Siemens Sonata/Trio, Erlangen, Germany). Different schemes of flip angle variations were implemented in a common TSE-sequence (MTX=256x208, FOV=(30cm)2, ESP=6.6ms, slth=5mm). In addition to the conventional modes with constant flip angles (180°, 90°, 60°), the mechanisms “hyperechoes” and “TRAPS” (both with flip angles 90° and 60°) were implemented. Simple linear and optimized sinusoidal flip angle variations were used for TRAPS. Comparative measurements between single- and multi-slice acquisitions (13 slices, dist=2.5mm) for two different echo train lengths (ETL=17 and ETL=27) were performed for MT investigations. Signal intensities were evaluated in ROI's of cerebral WM, GM, and orbital fat. The relative MT was assessed as the ratio of multi- to single-slice image intensities (MTrel). Results: Fig.1 demonstrates that the TSE180° displays a strong MT-effect. In comparison, the TSE90° and TSE60° show a largely reduced MT-effect. Hyperechoes and TRAPS display a MT-effect which is lower than for the TSE180°, but higher than for the corresponding conventional TSE. No significant difference between the different TRAPS-sequences was observed. The MT behavior of 3T is similar (Fig.2), but due to the lower number of slices achievable with TSE180° within the SAR limits the MT-effect of this sequence is artificially low. Generally, a slightly but not significantly stronger MT-effect for ETL=27 was seen (not shown). Discussion and Conclusion: The MT-effect of the different multispinecho-sequences is strongly correlated with the sum over all flip angles applied. Therefore, images acquired with the hyperechoand TRAPS-principle show less MT-effect and thus less contrast dependence on the number of slices than regular TSE180°-sequences. Additionally, the SAR deposited can be reduced by as much as 75% [1,2]. 1. Hennig J, Scheffler K, MRM 46:6-12 2. Hennig J, Weigel M, Scheffler K, MRM 49:527-35

418 Decrease of signal intensity of normal myometrium after Sinerem administration: a new MR finding with potential advantages on T-staging of uterine carcinoma P. Paolantonio, A. Laghi, C. Miglio, D. Marin, M. Celestre, R. Passariello; Department of Radiology, University of Rome, Rome, ITALY. Purpose: To evaluate the potential diagnostic advantages of intravenous Sinerem administration on T-staging of uterine carcinoma. Methods and Materials: Twelve consecutive female patients referred for gynecological carcinomas, 10 with corpus uterine cancer and two with cervical carcinoma underwent MR scan of the pelvis on a 1,5 T magnet. All patients were enrolled in a phase III study still in progress for the evaluation of Sinerem in lymph node staging. GRE T2*W sequences(TR/TE/FA/matrix/acq. time:1800 msec/15 mesc/30/230x512/13,51 min) were acquired on axial and sagittal planes before and 24 hours after iv administration of 2,6 mg/kg of body weight of Sinerem (Guerbet, Paris, France). Image analysis were performed by consensus by two radiologists and it included qualitative and quantitative analysis of signal intensity (SI) of both normal myometrium and neoplastic lesions before and after Sinerem administration. S/N and C/N were calculated. Modification of T-staging in images obtained after Sinerem administration was also evaluated. Statistical analysis was performed using t-test (p<0.05). Results: Qualitative analysis showed a significant decrease of SI of normal myometrium after Sinerem administration leading to a better lesion conspicuity. Quantitative analysis showed a statistically significant difference between SI of myometrium on plain and Sinerem-enhanced MR images assessed by means of t-test. C/N between lesion and normal myometrium significantly increased following Sinerem administration. In four patients Sinerem-enhanced images provided additional information leading to more accurate T staging. Conclusion: Intravenous injection of Sinerem provides a significant decrease of SI of normal myometrium with higher C/N between neoplastic lesion and normal myometrium. Sinerem-enhanced MRI provides better evaluation of myometrial neoplastic invasion increasing the conspicuity of neoplastic lesions in patients with endometrial and cervical carcinoma. Further studies are necessary in order to realize the physiologic mechanism of myometrial uptake of Sinerem.

RF-Systems

S221

RF-Systems 419 Automatic tuning and matching of a small multi-frequency saddle coil at 4.7T R. Pérez de Alejo Fortún, Sr.1, C. Garrido Salmón2, P. Villa Valverde1, I. Rodríguez1, J. Vaquero3, J. Ruiz-Cabello1, M. Cortijo Mérida1; 1NMR Unit, Instituto de estudios Biofuncionales, Universidad Complutense de Madrid, Madrid, SPAIN, 2Unidad de RMN, Instituto de Física de São Carlos, USP, São Carlos, BRAZIL, 3Unidad de Medicina y Cirugía Experimental, Hospital GU Gregorio Marañón, Madrid, SPAIN. Introduction: Experiments at different frequencies have been suggested in MRI; e.g. the use of hyperpolarized noble gas and other gases. A multi-tuned RF coil design can be implemented by the introduction of varicaps [1, 2], eliminating the need for perturbing the electromagnetic environment of the coil. A new circuit design for automatically tuning and matching a small saddle coil is presented. Materials and Methods: Images were acquired using a Bruker Biospec 47/40 spectrometer (4.7T). The saddle coil allows excellent access to the animal located inside and provides easy automatic tuning. A balanced sample coil was implemented (See Figure 1) [3]. The DC voltages were controlled using a digital potentiometer and a double operational amplifier. The scheme assures high stability and good isolation.

Figure 2. Results: In rat lung studies, the behavior of the loaded coil Q (see Figure 2) is very similar in all the frequency range (lower than 5%). This is corroborated by the SNR in the central and both extreme slices. The coil allows a very satisfactory 90 and 180 pulse adjustment. Coronal and axial images of the rat lung (19F B and 3He C images were overlaid on 1H images A) were consecutively acquired within 4 minutes (see figure 3).

Figure 1. The algorithm for tuning the system consisted of: a) previous characterization of the tuned coil inside the magnet and b) testing of all potentiometer positions around the value resulting from the first process.

Figure 3.

RF-Systems

S222

Discussions and Conclusions: The optimal matching and tuning response of the probe includes the frequency of 3He, 19F and 1H at 4.7T. The most recent approach to perform multinuclear images was developed for working at 3T with a coil working at 1H and 19F nuclei frequencies [1]. The use of saddle coil minimizes the number of varicaps in the final configuration and avoids the inclusion of several DC biases allowing a noise reduction. The new design assures high stability and digital noise isolation; it also facilitates autonomous work after preprogramming without the need for introducing an expensive, highly isolated operational amplifier. The main advantages of the system are time saving and operation versatility, crucial aspects to avoid noble gas depolarization. References: 1. Muftuler LT, Gulsen G, Sezen KD, Nalcioglu O. J Magn Reson 2002, 155: 39-44. 2. Sugiura S. U. S. Patent 4602213, July 1986. 3. Murphy-Boesch J, Koretsky AP. J Magn Reson 1983, 54: 526532.

a good basis to start optimizations for array coils. References: 1. ConceptII TU Hamburg-Harburg 2. Visualization Toolkit, www.kitware.com 3. Camelia Gabriel, Phys. Med. Bio. 41, 1966 4. S. Junge, Seifert, F., Rinneberg, H., ESMRMB 19, 424 Figure 1: Rat model with coil (left) and calculated rhs component of the B1-field (right) of an optimized quadrature surface coil 200MHz

420

RF-receiver coil for lung imaging at 0.15 Tesla A. Benattayallah1, J. Owers-Bradely2, R. W. Bowtell2; 1Diagnostic Radiology, University Hospital, Philipps University, Marburg, GERMANY, 2Physics and Astronomy, University of Nottingham, Nottingham, UNITED KINGDOM.

Optimized rf-coils with simulations of rat anatomical models S. Junge; Mri, RF Development, Ettlingen, GERMANY. Introduction: Coil development guided by simulations at high fields for calculating the B1-field distribution in loads, like anatomical models, is very useful for improving and optimizing the coil performance. We present simulation techniques and a rat head anatomical model which are useable for simulations of rf-coils for mapping the absolute B1-fields and their optimizations. Subjects and Methods: We used moment methods (CONCEPT II [1]) to solve Maxwell's equations in three dimensions with surface coils and a rat model as load fixing the boundary conditions. Dielectric, lossy bodies are defined by the shapes fixing their boundaries with the material constants found in table 1. Tissue Avg. Permittivity Brain 56.064735 Skull 19.181932 Muscle 60.868004 Table 1. Material constants [3]

Conductivity [S/m] 0.507771 0.134992 0.764407

The coils are introduced by their surfaces with wires including the lumped elements. We used the main tissues muscle, bone and brain, which have an effect for the B1-field distribution in a rat head. Segmented rat MR-Images were used as basic input for the program package vtk [2] to build up triangulated surfaces for the different tissues. In a second step the number of triangles for all surfaces were decimated as much as possible under protection of preserving their topology. In the past the efficiency of the used methods and techniques and their comparison with measurements [4] were demonstrated. The rat anatomical model was used to make parameter studies for optimizing linear and quadrature surface coils for brain imaging. Results and Discussion: Different coil sizes from 5mm up to 45mm were simulated as curved shaped designs. The absolute B1field values of the right hand side (rhs) component of the results were compared and the field distribution over the whole rat brain was weighted for a most homogenous image (figure 1 right). We got the best result with a combination of circular coil and a butterfly coil on a curved shape with a diameter of 34 mm (figure 1 left). This result was compared with measurements of an existing coil with about the same dimensions. These first comparable results are

421

Introduction: The design of the rf coil plays an important role in determining the signal-to-noise ratio (SNR) and hence the quality of the image, particularly at low-field where the coil noise dominates the total noise. The purpose of this work was to improve the image quality compare to a commercial chest coil (CC), which is a saddle-solenoid qaudrature coil. A highly sensitive elliptical Helmholtz (EH) quadrature receive coil, have been constructed specifically to fit closely around the human chest and to allow registered acquisition of both 1H and 3He images. Methods: The EH qaudrature coil was constructed applying a combination of an identical pair of EH linear coils (Figure 1 shows the circuit). Investigations were carried out in a 0.15T permanent magnet (ICG) with a horizontal field direction (perpendicular to patient-axis). This require the used of coil geometry which is different to that used in conventional system. All proton images were acquired with a 2D-GE sequence. Parameters for phantom images were: 256*256, TR/TE: 200ms/14ms, and FOV 300mm and for the chest images: 128*128, TR/TE: 25ms/5ms, FOV 440mm, and slices 13mm. Additionally, 3He lung images were acquired with a single shot RARE sequence (inter-echo time 18ms, FOV 434mm, and echoes 64)1.

RF-Systems Results: Measurement of the quality factor, frequency shift and input impedance were performed on the workbench, while field distributions and SNR were evaluated in the scanner with a 3 litters cylindrical CuSO4 phantom (0.5g/l). Signal profiles cross images of the phantom (left to right) acquired with the CC and the EH quadrature coils are depicted in Figure 2, showing a clear improvement in MR signal without any homogeneity degradation (the sensitivity of the EH coil was 243% higher than the CC coil). In vivo lung images of a healthy volunteer (registered 1H and 3He images in transverse and coronal plane) acquired with EH quadature coil are shown in Figure 3.

S223

TX and RX radio frequency (RF) coils [1]. A receiver system, equipped with a single TX volume RF coil and a double channel, out-of-phase, RX volume RF coils was developed [2]. In the present work, we investigated whether a double channel RX system equipped with RF surface coils is effective for achieving high isolation at operating frequency close to 100 MHz. Methods and Results: The novel system is made of three identical circular loop (diameter 6 cm) RF coils, tuned to f0=100.01 MHz when isolated. The three RF coils were positioned perpendicular to each other and the center-to-center coil distance, between each pair, was 8.5 cm. This configuration permits a open volume access for the sample. Due to the simmetry of the system, two resonant modes were observed (f1=99.4 MHz and f2=101.6 MHz). The lower frequency mode presents a reasonable homogeneous B1 spatial distribution and it can be set to the desired Larmor frequency. This mode is used for high efficiency RF pulse transmission and FID signal detection. Two of the RF surface coils were connected in double channel RX configuration as previously described [2]. The attenuation and phase-shift values of the auxiliary RX channel were adjusted to achieve high isolation at f1. With in-phase RX (summation mode) the measured isolation showed two peaks (f1, f2) with average value of -8 dB. The isolation with out-of-phase RX (subtraction mode) showed a dip of about –66 dB at f1; and at f2 the isolation presented a peak of -19 dB. Conclusions: We have shown that the double channel RX equipped with surface coils achieves very high isolation at the Larmor frequency. Potential applications of this work are in NMR imaging of solid materials, porous media, and solvent diffusion in solids. Here restricted motion of the probed spins leads to broad lines and very-short relaxation times (100 µs or less). Electron Paramagnetic Resonance (EPR) imaging of paramagnetic probes with ultra-short (10 µs or less) relaxation times is also an important application of the receiver described here. 1. Chen, C. N., Hoult, D. I., Biomedical Magnetic Resonance Tecnology, IOP Press (1989) 2. Alecci M, et al, J. Magn. Reson. 130, 272 (1998)

423 Discussion: The EH coil seems to be a good choice for lung imaging, since its elliptical shape conforms more closely to the human chest contour. Such a coil offers an increase in the sensitivity without suffering from any signal non-uniformity. The EH coil had 180% increase in SNR over the CC coil because of their optimum region of sensitivity, which restrict the amount of anatomy contributing to noise2, showing a good lung and chest images. References: 1. C.McGloin et al. ISMRM, Glasgow (2001). 2. J.R.MacFall et al. Radiology, 200;553-558 (1996).

422 Design and testing of a high isolation radio frequency receiver system for pulsed magnetic resonance imaging C. A. Curto1, G. Placidi2, A. Sotgiu1, M. Alecci1; 1INFM, Dept. STB, Università dell'Aquila, L'Aquila, ITALY, 2INFM, Centro di Risonanza Magnetica, Università dell'Aquila, L'Aquila, ITALY. Introduction: MRI scanners capable of detecting nuclei with very short relaxation times (100 µs or less) require a transmitter (TX) and receiver (RX) system with “zero” dead time. The achievement of this condition depends mainly on the overall isolation between

Possibilities and limitations of inductive coil coupling in small sample MRI H. Graf, P. Martirosian, F. Schick; Section on Experimental Radiology, University Clinic Tuebingen, Tuebingen, GERMANY. Purpose/Introduction: To study the potential of inductive coil coupling for a sample size and field of view (FOV) optimized MR examination of small objects and animals in whole body MR scanners for humans. Subjects and Methods: Resonant circuits (RC) with solenoid coils of an inner diameter corresponding to a mouse/rat were built from copper wire for varying diameter, length and turns per unit length (N/l). The coils were tuned to the 1H Larmor frequency at 1.5 T with a trim capacitor. Signal and flip angle amplification (tau) as well as the improvement (eta) of the signal to noise ratio (SNR) in the inner of the coils was examined arranging the solenoids at the isocenter of the scanner inside the standard head coil (setup see Fig.1).

S224

RF-Systems 424 Theoretical comparison between implantable microcoils and external superconducting coils for rat brain localized NMR spectroscopy M. Armenean, O. Beuf, H. Saint-Jalmes; Laboratoire de RMN, UMR CNRS 5012, CPE, Université Claude Bernard Lyon1, Villeurbanne, FRANCE.

The examinations were done for water phantoms and for structured samples. Theoretical investigation according to the theory of a strongly non-ideal transformer was carried out. Results: Solenoids with an inductance < 1 µH are practically still tunable with capacitors of several pF at 63.8 MHz. For a coil volume corresponding to a mouse the limit for N/l lies at about 0.7/cm. For such coils flip angle and signal amplification values of about 30 were found, whereas the SNR improvement ranged between 2 and 6. With increasing coil diameter N/l or l has to be reduced. At enlarged load flip angle and signal amplification reduced clearly and signal amplification and SNR improvement became more and more equal at a value of about 2 to 4. This behavior qualitatively could be explained by the theoretical calculations (see Fig.2) taking into account the additional electronic noise with the use of the adapting coils.

Figs.3a/b show images from a dead mouse obtained at the same measurement protocol using a standard finger coil (a) and using a RC in the head coil (b), Figs.3c/d show images from a dead rat obtained at the same protocol without (c) and with the use of a RC (d).

Discussion/Conclusion: Inductively coupled solenoid coils offer within the described limits a simple and low cost possibility to adapt whole body MR scanners to the examination of small objects. The gain in SNR either can be used to reduce the measurement time at fixed spatial resolution, or, especially with scanners supplied with high performance gradient systems, to perform examinations at higher spatial resolution.

Purpose/Introduction: During different sleep phases important metabolites variations are localized in very small brain area named “activation centers”. The quantification of these metabolites is important to understand the mechanisms of some brain diseases. In this communication it has been compared by electromagnetic simulations the signal to noise ratio (SNR) of two NMR spectroscopy techniques: one using external superconducting radiofrequency (RF) coil and the other, implantable cooper microcoil, both as reception coils only. In our study we considered that the activation center has a volume of 0.5µl and is located in the center of the rat brain. Subjects and Methods: The superconducting coil with a diameter of 40mm placed around the rat head has been considered. The dimensions of microcoil placed in the center of the activation center are: 1000µm length, 500µm width, four turns, 20µm interturn spacing, 38µm wire thickness and 22µm wire width. The rat brain is approximated by a sphere of 30mm diameter, 1S/m conductivity and 81 relative permittivity. The simulations were performed at a frequency of 85.13MHz. If the activation center is not selected with gradients, the coils quality is given by the useful over total detected signals ratio, α. In the case when the activation center is selected with gradients the discrimination factor between coils is β=volume integral(B1/R1/2dV), where B1 is the RF magnetic field produced by a unit current, R is the electrical coil resistance (includes both magnetic and electric losses) and V is the activation center volume. The β factor gives the SNR. Results: The calculated inductance and resistance are 12.8nH and 0.424Ω respectively for the microcoil and 77.3nH and 0.107Ω respectively for the external superconducting coil. The electrical resistance of the superconducting coil is only due to the losses in the conducting environment (brain). The discriminative parameters without activation center selection are α=3.7×10-5 for the external coil and α=0.61 for the implantable microcoil. In the same time with activation center selection the β factor is about one hundred times bigger for the implantable microcoil than the external coil. Discussion/Conclusion: The implantable microcoils placed closed to the activation center allow the separation between the signals from the activation center and surrounding volume. For external coil the useful signal is completely hidden by the signals from the entire brain volume. In the same time the SNR is much better for the implantable microcoils than for external coils even superconducting.

425 Design of Figure-of-Eight Radio Frequency Surface Coils for MRI at 1.5 T M. Alfonsetti1, G. Placidi2, A. Sotgiu1, M. Alecci1; 1INFM, Dept. STB, Università dell'Aquila, L'Aquila, ITALY, 2INFM, Centro Risonanza Magnetica, Università dell'Aquila, L'Aquila, ITALY. Introduction: Figure-of-eight (FO8) RF coil comprises current flowing along two semicircular elements and along two central linear elements [1-2]. In the central region the B1 field is perpendicu-

Imaging: Sequences and Techniques lar to the linear currents. It was shown that FO8 coils with radius R=9 cm exhibit higher B1 amplitude along the coil z-axis, as compared to the circular loop (CL) coil [3]. We present a theoretical and experimental analysis of the optimal design parameters of FO8 coils obtained in a wide range of geometrical parameters at 1.5 Tesla. Methods and Results: The B1 components were calculated by numerical integration of Biot-Savart law (I=1A). To study the optimal design of FO8 coils (Fig. 1) we estimated the: peak field, coil axial sensitivity and axial selectivity versus the radius (28 cm) there is a wide range of 2s values which gives a sensitivity gain. On the contrary for small radii the sensitivity gain is obtained only for 2s<2 cm. The axial selectivity increases as R increases and it reaches an asymptotic value depending on 2s. The axial position of the peak B1 field shows only a small increase as R increases. The measured B1 values of sensitivity, selectivity and maximum position are in good qualitative agreement with theory, but the measured sensitivity values are lower than the expected values, particularly at small 2s values. Conclusions: A significant sensitivity improvement is obtained with FO8 coils, and by selecting the R and 2s values the B1 selectivity and axial peak position can be adjusted within few cm. The FO8 coil shows benefits for signal optimisation over specified VOI in MR spectroscopy and imaging. 1. Smith MA, et al, MRI 4, 455, 1986. 2. Seton HC, et al, MAGMA 1999,8,116-120. 3. Alfonsetti M, et al, INFMeeting 2002, 7, 2002. 4. Maier LC, et al, J. Appl. Phys. 23, 68, 1951.

S225

Imaging: Sequences and Techniques 426 Un-prepped MR colonography in paediatric patients with inflammatory bowel disease (ibd) P. Paolantonio1, A. Laghi1, M. Celestre1, D. Marin1, S. Cucchiara2, R. Passariello1; 1Department of Radiology, University of Rome, Rome, ITALY, 2Department of Pediatrics, University of Rome, Rome, ITALY. Purpose: To assess the feasibility of MR colonography performed without colonic cleasing by means of fecal tagging using oral administration of barium sulfate in a pediatric population with inflammatory bowel disease. Subjects and Methods: 10 paediatric patients (mean age 13.8 yrs) with known IBD assessed at colonoscopy with biopsy were enrolled in the study. Fecal tagging was performed by the oral administration of a fixed amount of 150 ml of highly-concentrated barium sulfate at major meals starting three days before MR examination and with a high iron content diet. Immediately before the examination the colon was distended by a water enema. MR imaging protocol included T2w HASTE (TR, inf; TE, 90msec; FA, 180°; acq time, 20 sec) and contrast enhanced T1w 3D VIBE (TR, 5,2 msec; TE, 2,6 msec; FA, 20°, acq time, 18 sec) and T1w 2D turboFLASH (TR/TE/FA/TI: 8,5 msec, 4 mesc, 10°, 10 msec) sequences. Contrast enhancement was obtained by i.v injection of 0.1 mmol/kg b.w. ml of Gd-chelate. Results: Image quality was judged to be optimal in all patients. Fecal material was completely tagged in all the cases and residual luminal air did not cause any significant imaging artefact. Patient tolerance was high. All the patients showed signs of IBD, Crohn's disease in three cases and ulcerative colitis in the remaining seven cases. MR correctly defined thickening of the colonic and localization and extension of the inflammed bowel wall, by means of contrast enhancement following Gd-chelate injection. Conclusions: Unprepped MR colonography by barium fecal tagging is a promising and well tolerated diagnostic method in the evaluation of colonic involvement in pediatric patients with IBD.

S226

Imaging: Sequences and Techniques

427 Evaluation of the new i-PAT head coil and partially parallel acquisition technique (GRAPPA) in cerebral imaging: A volunteer study J. Tintera1, J. Gawehn2, T. Bauermann2, P. Stoeter2, P. Stoeter3; 1Department of Radiology, IKEM, Prague, CZECH REPUBLIC, 2Institute of Neuroradiology, University Clinic, Mainz, GERMANY, 3Institute of Neuroradiology, University Clinic, Prague, CZECH REPUBLIC. Introduction: Partially Parallel Acquisition (PPA) technique uses information from an array of coils to facilitate reconstructions either in k- or image space. After first successful applications have been published in cardiac and pulmonary vascular imaging, we tested PPA GRAPPA (1) in an 8-channel head coil for various cerebral imaging. Methods: The new i-PAT head coil consists from 8 parallel surface coils. This coil was implemented in 1.5 T Sonata scanner (Siemens, Erlangen). Standard and GRAPPA mode with various acceleration factors (AF) and number of reference lines (RL) were varied to test the image quality. Some results were also compared with a standard head coil. The following imaging techniques were tested in volunteers: 1) Intracranial Time-Of-Flight (TOF) MRA: 3D-FISP (TR=40ms, TE=7.15ms, flip angle (FA)=25°, voxel=0.8x0.4x1mm, 80 slices) and a) without PPA (acquisition time (TA)=9min 48s) b) with AF=2, RL=40 (TA=6 min 51s (70%)), c) AF=3, RL=64 (TA=4 min 53s (50%)). 2) 3D high resolution imaging: MP-RAGE (TR=1900ms, TI=1100ms, TE=3.93ms, FA=15°, voxel=1x1x1.25mm, 72 slices) was performed with a) standard head coil and with b) i-PAT coil. 3) Cerebral activation (finger tapping) with BOLD contrast: GE EPI (TR=6 s, TE=59 ms, FA=90°, bandwidth=1300 Hz/pixel, voxel=1.8x1.8x4, 27 slices) was measured with a) head coil, with b) i-PAT coil and with c) i-PAT coil with GRAPPA, AF=2, RL=64. Results: 1. There is no visible loss of TOF MRA quality, mainly of peripheral artery “enhancement” using sequence 1b) in spite of an important time advantage. Further reduction of measurement time as in condition 1c) leads to sever image quality impairment (fig.1).

2. Although some cortical areas clearly showed better differentiation of the cortex from the subcortical white matter using i-PAT coil, this improvement of image quality was confined only in the brain surface. 3. An average level of standard deviation (STD) in the iPAT coil images was found as 0.82 of that measured with standard head coil. The use of GRAPPA (3c) leads to SNR improvement only in the image periphery where STD was about 0.7 of 3b). On the contrary, a higher level of artifacts decreased SNR in

the image center. Slightly higher noise in this region probably caused the reduction of the activated area in fMRI using GRAPPA mode (Fig.2)

Discussion: Our testing of the i-PAT coil in combination with the GRAPPA mode revealed some advantages but also some limitations for its routine application in neuroradiology. 1. Griswold MA et al. [2002] Magn.Reson.Med 47:1202-1210

428 Cardiac cine parallel imaging on an open 0.7T MRI apparatus M. Takizawa1, T. Goto1, M. Nonaka2, H. Mochizuki2, S. Nagai2, H. Takeuchi2, Y. Taniguchi3, H. Ochi3, T. Takahashi1; 1Research & Development Center, Hitachi Medical Corporation, Chiba, JAPAN, 2MRI Design Department, Hitachi Medical Corporation, Chiba, JAPAN, 3Central Research Laboratory, Hitachi Ltd., Tokyo, JAPAN. Introduction: Parallel imaging is valuable for cardiac imaging on tunnel shape MRI apparatus. But there are few reports on open MRI. So, this study realizes cardiac cine parallel imaging on an open 0.7T MRI. Imaging time shortened in any slice direction by using a dedicated four channel RF receiving coil consisting of solenoid coils and butterfly coils. Subject and Method: Coil shape was designed by using RF coil simulation under consideration of biological load. Simulation study showed that image quality strongly depended on coil shape and imaging plane, and optimal shape consisting of crossed two solenoid coils and twisted two butterfly coils was found. Modified image-domain reconstruction algorithm was equipped with auto-calibration on a 0.7 T open MRI apparatus (Altaire(R), Hitachi Medical Corp.). Cine images were obtained with Balanced SARGE sequence. Image quality was evaluated by using cylindrical phantoms and 5 healthy volunteers. Results: Multi-slice phantom images showed no visible artifacts. Breath hold cine images with acceleration factor of 2 were evaluated carefully. With and without auto-calibration, there were no visible unfolded artifact and/or motion artifact. Thus parallel imaging improved short axis, long axis and four chamber images in acquisition speed, time resolution, and/or spatial resolution. Figure shows the conventional (A) and parallel (B) image on four chamber view. Conclusion: Dedicated RF coil enabled cardiac cine parallel imaging on open MRI.

Imaging: Sequences and Techniques 429 "Zooming" with SENSE: a 4 channel phased array coil to locally explore brain cortex C. C. Meca1, E. Giacomini1, P. Scifo2, D. Le Bihan1, F. Lethimonier1; 1Unaf, SHFJ/CEA, Orsay, FRANCE, 2Nuclear Medicine, Scientific Institute H San Raffaele, Milan, ITALY. Purpose/Introduction: Through partial phase encoding, parallel imaging (eg ASSET/SENSE or SMASH) reduces acquisition times and therefore enhances temporal/spatial resolution and/or minimizes sensibility to susceptibility and eddy current artefacts1, especially when used with EPI. Parallel imaging methods rely on the information encoded in the sensitivity profile of each loop of the phased array coil to regain the complete unaliased image. The loop element number is, however, limited by the receiver number of channels and is seen as a limiting factor for SENSE acquisitions. In this study we developed a dedicated 4 channel SENSE coil demonstrating the feasibility of using SENSE with a low number of small coils elements to locally explore cortex regions with high resolution and reduced artefacts. Subjects and Methods: All experiments were made on a 1.5T Signa LX scanner (GEMS, Milwaukee, USA). The developed “zoom” coil consisted of 4 regular resonating rectangular loops (5x13cm) aligned by their longest side on a cylindrical support. The decoupling of adjacent coils was achieved through the use of high impedance preamplifiers and inductive negative transformers between the loops2. Coil overlapping, although vastly used in phased array coils, was not used as it increases the geometric factor, g, and reduces SNR in reconstructed images2. SENSE EPI diffusion tensor (reduction factor 2), 128x128, TE/106ms, TR/2000ms FOV/24cm (1.9x1.9x5mm voxel size) with b values [0, 800]s/mm2 and 21 directions, were acquired to assess the possibilities of the developed coil. Results: Diffusion EPI images were obtained and ADC and fractional anisotropy maps were calculated. With SENSE acquisition, susceptibility artefacts and eddy currents are not visible in the calculated diffusion images. High contrast fractional anisotropy maps are obtained and the values for ADC and FA are coherent with those obtained with standard SENSE acquisitions.

S227

Discussion/Conclusion: SENSE minimizes phase artefacts (eddy currents distortions, ghosting or susceptibility artefacts) by reducing the Echo Train Length, as shown in fig.1. We evaluated the use of small, dedicated, SENSE surface coil arrays to allow the exploration of cortex regions with high SNR3, high spatial resolutions, and reduced artefacts. The use of SENSE to explore brain cortical regions will permit to acquire high-resolution acquisitions images on limiting conditions such as low number of channels available, or difficult whole brain acceptable shimming (very high field systems). References: 1. Pruessman KP et al., MRM, 42: 952-962 (1999) 2. Weiger M et al., MRM, 45: 495-504 (2001) 3. Meca CC et al, 10ème GRAMM, Angers 2003

430 Imaging tumor tissue during histological processing: 1Hethanol MRI M. O'Brien, D. J. O. McIntyre, J. R. Griffiths; CR-UK Biomedical Magnetic Resonance Research Group, St George's Hospital Medical School, London, UNITED KINGDOM. Purpose/Introduction: It is often necessary to compare in vivo tumor MRI images with sections of the resulting histological specimens. We have imaged specimens through the first stages of classical histological preparation: formaldehyde fixation and de-watering in graded ethanol solutions. Surprisingly, although the final specimens contained negligible water as judged by MRS, satisfactory MRI was possible from the ethanol signal. Acquired volumes were assessed for changes in size, information loss and distortion. Subjects and Methods: All image volumes were acquired on a VARIAN UNITY INOVA 4.7T system using a FAST GE3D sequence, obtaining a 1283 volume with TR 20ms, TE 10ms and 4cm3 FOV. GH3 prolactinomas were imaged, excised and fixed in formalin. Fixed tumors were imaged, washed and de-watered for 24 hours in each graded ethanol solution (30%, 70% and 100%) prior to imaging the ethanol-infused tissue. Maximum signal amplitude, mean signal amplitude and standard deviation were recorded for a VOI of 163. The coefficient of variation was used to give a comparable dimensionless number representing contrast. Changes in size and distortion were assessed by co-registration using non-scaled principal axes transformation1 software written in IDL 5.4. Results: The formalin-infused specimen showed greater contrast than was observed in vivo . The coefficient of variation was lower in 30% and 70% ethanol, rising again in 100% ethanol (figure 1). Through the fixation and drying process, co-registration of the volumes showed no change in tumor volume and no distortion of tissue mass (figures 2, 3). 1H MRS of the specimen in 100% ethanol showed negligible water signal, thus the MRI was acquired from the ethanol signal.

S228

Imaging: Sequences and Techniques 2t, and the gradient during this time has negative amplitude. The second RF pulse has a flip angle of π/3 for the negative sequence, and 2π/3 for the positive [2]. The characteristic length scale of an iMQC scan is defined as the correlation distance [1] given by dc = π/γγGT where G is the gradient amplitude and T its duration. Three different correlation distances (dc= 266, 133, and 67µm) were examined to ensure that the scans behaved the same over a range of parameters. Figure 1. +DQC pulse sequence.

Discussion/Conclusion: Formalin "fixes" cellular structures by forming polymeric chains; these may account for the enhanced MRI contrast . The contrast observed from ethanol imaging of the de-watered specimen is affected by the removal of lipid and by the interaction of ethanol with macromolecules. A major factor is the exclusion of ethanol from the fixed cell cytoplasm2, restricting it to the interstitial space. The ethanol proton density will therefore vary more with tissue cellularity than is the case with water in living tissues. References: 1. Alpert N M, Bradshaw J F and Kennedy D [1990] J.Nucl. Med. 31:1717-22 2. Pearse AGE. [1980] Histochemistry Theory and app. 53-56 This work was supported by Cancer Research UK

431 Intermolecular positive and negative double quantum coherences in the human brain D. P. Bulte1,2, M. D. Noseworthy1,2; 1Diagnostic Imaging, The Hospital for Sick Children, Toronto, ON, CANADA, 2Department of Medical Biophysics, University of Toronto, Toronto, ON, CANADA. Purpose/Introduction: Intermolecular multiple quantum coherences (iMQC) [1] have been used in MRI to probe tissue structure on scales beyond those accessible by conventional scans. Theory implies that intermolecular positive double quantum coherence images will be identical to negative double images [2]. Such a comparison is not possible with zero-quantum coherences [3]. However, in practice positive double quantum coherence scans prove to reliably give higher signal intensities in vivo using clinical hardware. Subjects and Methods: Using a GE Signa 1.5T LX-MRI, positive double quantum coherence (+DQC) imaging of a single axial slice through the brain lateral ventricles was compared with negative double quantum coherence (-DQC) imaging. The positive pulse sequence is shown is Figure 1. The negative sequence is identical except the time between the second RF pulse and the 180° is TE/2 -

Results: Figure 2. shows +DQC and, -DQC scans, and their %difference, with dc=133µm. As shown, the percentage change in signal over much of the slice is approximately 30%, but is as much as 80% in the highly vascular regions. Similar percentage changes were apparent at correlation distances of 266 and 67µm.

Figure 2. Discussion/Conclusion: To date a comparison between positive and negative double quantum coherence techniques has not been evaluated. Based on quantum mechanical theory the +DQC and -DQC images should be identical. We suggest iMQC is not necessarily straight forward to evaluate in human subjects with a clinical scanner. Normal amplitude-modulated, non-adiabatic pulses, as used by the sequences shown here, can result in dramatic differences between positive and negative iDQC images. We are evaluating new phase cycling schemes, rf pulse design, and B0 homogeneity in attempt to reach the theoretical equivalence between these two approaches. References: 1. Warren WS, et al. (1998) Science, 281:247-251. 2. Zhong J, Chen Z, Kwok E. (2000) JMRI. 12:311-320. 3. Rizi RR, et al. (2000) MRM. 43:627-632.

432 A multi-echo phase contrast gradient echo sequence for accelerated MR elastography S. Maderwald, K. Uffmann, C. J. Galbán, A. de Greiff, S. Mateiescu, M. E. Ladd; Departement of Diagnostic and Interventional Radiology, University Hospital Essen, Essen, GERMANY.

Imaging: Sequences and Techniques Introduction: The mechanical properties of soft tissue can be determined from an MRI sequence modified to visualize mechanical shear waves propagating through the tissue [1]. This technique, known as MR elastography (MRE), currently suffers from long scan times. The purpose of the present study was to implement a multi-echo phase contrast (PC) gradient echo sequence and compare it to a conventional single-echo sequence. Methods: Mechanical waves were introduced in agar-agar gel phantoms by a piezoelectric actuator [2]. A multi-echo PC FLASH sequence was implemented with motion-sensitizing gradients (MSG) synchronized to the mechanical vibration. All images were acquired with a 1.5T SIEMENS Sonata system. Typical scan parameters were: TR 40 ms; matrix 256x256; FOV 200x200 mm2; flip 15°; MSG cycles 1; MSG amplitude 25 mT/m; mechanical frequency 200 Hz. The echo train length (ETL) was varied from 1 to 4. TE was set to the lowest possible value to ensure the highest SNR for a given sequence: 8.95 ms, 11.3 ms, 10.7 ms, and 10.5 ms for ETL 1-4, respectively. The single-echo sequence was used as the standard of reference. Images with higher ETL were compared regarding SNR and artifacts. Profiles through the shear waves in the phase images were analyzed in MATLAB to determine the wavelength and calculate the elasticity. Results: The time reduction from the multi-echo sequence was nearly proportional to ETL. Scan times were 22 s, 11 s, 7.4 s, and 5.8 s for ETL 1-4, respectively. The measured wavelength was nearly identical for all ETL (44 mm, 43 mm, 43 mm, 43 mm for ETL 1-4). The calculated shear modulus was thus also independent of ETL (81.4 kPa-85.2 kPa), which correlates with published elasticity results for 2.5% agar-agar gel [3]. Ghosting artifacts in the images increased slightly with increasing ETL factor. Discussion: These results demonstrate that a multi-echo sequence is suitable for MRE, since it provides equivalent elasticity values in significantly reduced time. Investigation of higher ETL may lead to even higher acquisition speed. A single-shot EPI sequence would offer the shortest scan times, but at the cost of unacceptable artifacts. The ETL must be chosen to a value which offers a compromise between imaging speed and phase image quality. References: 1. Muthupillai R, et al., MRM 36:266-274 (1996) 2. Uffman K, et al., MR Engineering, 15(4) :239-254 (2002) 3. Hamhaber U, et al., MRM 49:71-77 (2003)

433 A 3D MDEFT sequence for anatomical brain scans R. Deichmann, R. Turner; Wellcome Department of Imaging Neuroscience, Institute of Neurology, London, UNITED KINGDOM. Introduction: MDEFT is a well-known sequence for anatomical imaging at high field strengths [1]. The applicability at 1 Tesla has likewise been shown [2]. Purpose of this work is the optimization of a 3D MDEFT sequence [3] for anatomical brain imaging with whole brain coverage and an isotropic resolution of 1 mm. Theory: The matrix size is fixed to 256x224x176 (Read/Phase/Partition). The sequence structure is: 90deg - t1 180deg - t2 - AQ. AQ comprises the acquisition of 88 3D phase encoding steps in a segmented centric order, using a spoiled FLASH readout with a repetition time TR and B1 insensitive RF pulses [4] with a flip angle FA. This is repeated 2*224 times with different 2D phase encoding. The optimization starts with trial values for t1 and t2. From the total duration (fixed to 12 min), TR and the readout bandwidth BW are calculated. The maximum possible FA not caus-

S229

ing contrast smoothing is obtained by calculating the magnetization evolution during AQ and the differential point spread functions [5]. t1 and t2 are optimized for maximum CNR (white/gray matter). During t2, arterial spin tagging and fat saturation may be applied to reduce signal from vessels and scalp. Methods: Scans were performed on a 1.5 T scanner (Siemens Sonata), using an 8 channel phased array head coil. The optimum parameters for this field strength are: t1/t2=220ms/310ms, TR/TE/FA=12.2ms/3.6ms/23deg, BW=106Hz/Px. The sequence was tested on a healthy volunteer. For comparison, the scan was repeated without fat saturation and spin tagging, and using a standard head coil. Results: The CNR is considerably higher than values reported for an optimized MPRAGE sequence at 2T [5], resulting in excellent white/gray matter segmentation. Even the measurements performed with the standard head coil show a clear improvement. Omission of spin tagging causes moderate pulsation artifacts, especially in the corpus callosum and the brain stem. Omission of fat saturation yields a high scalp intensity which may cause segmentation algorithms to fail. Discussion: The optimized 3D MDEFT sequence allows for high contrast anatomical brain scanning with an isotropic resolution of 1 mm within 12 min. The use of fat saturation and spin tagging pulses improves image quality considerably. References: 1. Ugurbil et al. (1993) MagnResonQuart 9:259-277 2. Norris et al. (1997) 5th ISMRM Meeting, 686 3. Lee et al. (1995) MRM 34:308-312 4. Deichmann et al. (2002) MRM 47:398-402 5. Deichmann et al. (2000) NeuroImage 12:112-127

434 K-space reordering to reduce the flow sensitivity of balanced SSFP K. Scheffler1, J. Leupold2, O. Bieri1; 1Medical Radiology, MRPhysics, Basel, SWITZERLAND, 2Radiology, Medical-Physics, Freiburg, GERMANY. Introduction: Several MRA methods have been proposed that are based on the balanced SSFP technique, for example time-of-flight (TOF) or contrast enhanced (CE) MRA (1,2). However, the very rapid flow in major arteries can disturb the coherent phase relation that is required for the balanced SSFP technique. In this contribution we present an examination on the flow sensitivity of balanced SSFP, depending on the chosen k-space reordering and other parameters. Methods: Two sources of flow sensitivity have to be taken into account: flow-induced dephasing between excitation and echo refocusing, and flow-induced dephasing between consecutive excitation pulses. The signal decay that is induced between consecutive excitations is more critical. For 2D sequences the slice and the read direction is intrinsically flow compensated, whereas the phase encode direction is not flow compensated. Therefore, flow along read or slice direction does not disturb the steady state and the full SSFP signal can be detected. For 3D sequences both phase and slice direction are not flow compensated (i.e. there is a nonzero first moment between excitation pulses). A flow-induced dephasing of only 180° or more between excitation pulses can result in a signal decay of about 90% which makes balanced SSFP quite sensitive to motion or flow along phase direction and slice direction (only 3D). Results: Flow along the read direction did nearly show no signal

S230

Imaging: Sequences and Techniques

reduction for flow velocities up to 0.5 m/s. For the 3D measurement a strong (periodic) signal decay could be observed for flow along phase encode and along slice select direction. This periodicity was also predicted by our simulations and was also derived by Patz (3). For higher flow velocities (from 0.2m/s) the periodic signal decays to a very low static signal that was comparable to the signal of a spoiled FALSH sequence. Discussion: The non-zero first moment leads to significant flow voids along phase encode and slice select direction. The use of balanced SSFP for TOF or CE MRA is therefore limited. However, compensation of the first moment along slice and phase direction can compensate these signal drops. Another possibility is to chose reordered k-space trajectories to minimize flow induced dephasing along phase encode directions.

435 Neuronavigation benefits from T2 weigthed 3D-Turbo-SpinEcho (TSE) images J. Gralla1, R. Guzman2, L. Remonda1, C. Kiefer1; 1Neuroradiology, University Hospital Bern, Bern, SWITZERLAND, 2Neurosurgical Clinic, University Hospital Bern, Bern, SWITZERLAND. Purpose: The clinical experiences with the application of an isovoxel 3D turbo spin echo sequence modality for neuronavigation and functional imaging are presented. Considering as example the infiltrate low grade glioma and the configuration of cystic lesions the benefits for neurosurgical planing are evident. Subjects and Methods: A 3D-TSE sequence with a turbo factor of N=25, S=15 slabs with 10 slices was used. The number of concatenations was set to NC = 2 to avoid saturation effects at the transitions of the slabs. With a repetition time of TR = 2570 ms and an echo time of TE = 106ms the measurement takes TA=12:27 min:sec. For later registration purposes 6-10 disposal skin fiducials has been tagged to the head of the patient. The navigational system (Stealth Station, Medtronic, Broomfield, CO, USA) is capable of matching and multiplanar reconstructing the measured dicom images. Results: The excellent contrast on the T2 weigthed 3D-TSE images makes it easy to identify the affected regions. The registration accuray (root mean square error) is below 2mm which is acceptable for neuronavigation. Anatomical landmark checks were in all cases satisfying. Conclusion: It was shown that neuronavigation benefits from including T2 weigthed 3D-Turbo-Spin-Echo data sets. The spatial accuracy which is necessary for intraoperative visualization is within clinical acceptance (< 2mm).

436 A method for interleaved acquisition of vascular input function for DCE-MRI in experimental tumours in rats D. J. O. McIntyre, C. Ludwig, A. Pasan, J. R. Griffiths; CR-UK Biomedical Magnetic Resonance Research Group, St George's Hospital Medical School, London, UNITED KINGDOM. Introduction: Derivation of Ktrans from DCE-MRI data ideally requires an arterial input function (AIF) from each subject, obtained simultaneously with the tumour signals. We describe here a method for obtaining vascular [Gd-DTPA] concentration in rat from the large tail vessels, interleaved with surface coil tumour data.

Methods: MRI was performed on a 4.7T horizontal bore Varian Unity Inova. Tail images were acquired using a 24 mm long by 8 mm diameter solenoid. Tumour data were acquired using a threeturn 25mm ID coil. A spectrometer-controlled PIN diode switch was used to switch between the coils. GH3 prolactinomas were grown in Wistar Furth rats. AIF data alone were obtained from 3 rats and interleaved AIF/tumour data from 2 rats. T1-weighted gradient echo images (TR=100, TE=4, FOV=1cm, 64x64 points, 6mm slice thickness, 6.4 second time resolution) were acquired from the tail. For the DCE run, the signal from the tail was saturated by slice-selective 90º pulses followed by crusher gradients, followed by a 40ms delay prior to acquisition, eliminating inflow effects for velocities up to 30 cm/sec. Pre-injection data were obtained with up to 32 scans to maximize SNR. 0.1mmol/kg Omniscan was injected as the scan started. 140 images were acquired post-injection. [Gd-DTPA] calculations were based on an assumed value for baseline plasma T1 as inflow effects preclude the use of long delays in IR or SR images. Analysis software to calculate [Gd-DTPA] and Tofts Ktrans was implemented in Matlab 7. Bi and triexponential fits to vascular concentrations were performed using Origin 4.1 Results: No inflow effects were observed in the veins, while the pulsatility and high velocity of arterial flow caused ghosting of the arterial signal. Since the time resolution is insufficient to resolve the first pass of contrast agent, the venous and arterial concentrations will be similar. [Gd] was calculated from the mean values of pixels lying within the two tail veins, and the arterial pixels were not used. The curves obtained closely resemble literature values (Figure 1). Biexponential fits to the AIF data were used to calculate maps of Ktrans and ve by the method of Tofts et al. (1) (Figure 2). Conclusion: PIN diode switching and a saturation recovery imaging method allow interleaved tail vessel AIF and tumour data to be acquired reliably with high sensitivity and good time resolution in a rat model. References: 1. PS Tofts, AG Kermode Magn Reson Med (1991) 17:357-367

Imaging: Sequences and Techniques

437 Spatial encoding using readout segmentation (SPURS): An alternative sampling scheme for single-shot, multiple spinecho MRI D. A. Porter, E. Mueller; MR Applications Development, Siemens Medical Solutions, Erlangen, GERMANY. Introduction: Single-shot sequences, such as RARE (2), HASTE (3) and GRASE (4), use multiple RF refocusing pulses to reduce the susceptibility artefacts seen with EPI (1). Of these, only GRASE has a small enough number of refocusing pulses to allow slice-selection to be used, but image quality can be degraded by T2-Decay and off-resonance effects, particularly when the number of refocusing pulses is very low. This paper considers an alternative approach to data sampling, providing a significant reduction in susceptibility artefact compared to EPI, but requiring less RF refocusing pulses than GRASE. The reduced number of refocusing pulses should be of value in reducing SAR, improving the slice-profile and limiting artefacts in diffusion-weighted imaging. Subjects and Methods: A series of EPI readouts are performed using 180o pulses to generate a spin-echo at the centre of each EPI echo-train (see figure). Each EPI echo-train samples a segment of k-space along kx and the entire range of k-space along ky. The reduced kx range results in a substantially shorter echo-spacing than with EPI. The sequence was implemented on a Siemens Magnetom Sonata system (40mTm-1, 200Tm-1s-1) and images were acquired from a healthy volunteer using the following protocol: FOV 173x230mm; matrix 96x126; slice thickness 5mm; 7 spin-echoes with spacing 27ms; EPI echo-train length 96 and echo-spacing 0.24ms; echo-time 108ms.

S231

Results: Volunteer images demonstrate the capability of generating good quality, single-shot T2-weighted images. Discussion: Because all ky points are sampled consecutively at each spin-echo, SPURS avoids the periodic k-space signal modulation seen with GRASE. The trade-off for this smoother ky signal variation is an amplitude variation along kx due to T2 decay between spin-echoes; this is equivalent to the ky signal variation seen with multi-shot Turbo-Spin-Echo (TSE) sequences and is associated with a relatively benign type of image artefact. Contrast behaviour is also similar to TSE as the central region of k-space is acquired at a single echo-time. SPURS requires a high gradient slew rate, but uses a relatively low gradient amplitude. This combination is convenient for whole-body gradient coils, as stimulation considerations often limit the available slew rate at high gradient amplitudes. References: 1. Mansfield P, [1977] J. Phys C. 10:L55-L58. 2. Hennig J et al, [1986] Magn. Reson. Med. 3:823-833. 3. Kiefer B et al, [1994] J. Magn. Reson. Imaging. 4(P):86. 4. Oshio K, Feinberg DA, [1991] Magn. Reson. Med. 20:344349.

438 Ultrashort TE Imaging at 3T C. Schroeder1, P. Börnert1, J. van Eggermond2; 1Tomographic Imaging Systems, Philips Research, Hamburg, GERMANY, 2MR Systems, Philips Medical Systems, Best, NETHERLANDS. Introduction: MR imaging of short T2 species has been a topic of continuous research [1]. Short T2 components appear in highly ordered tissues such as tendons, ligaments, cartilage and bone, i.e., tissues that exhibit low signal contents in conventional imaging sequences. To image short T2 species, sequences employing ultrashort echo times (UTE) are required. Such sequences may benefit from high field strength, yielding an improved signal-to-noise ratio and spatial resolution. In this work, UTE imaging at 3T is investigated in combination with a fast spiral/TwiRL read-out. Theory and Methods: To achieve ultrashort echo times, half-sinc RF excitation pulses are used, limiting the echo time to only the hardware switching time [2]. For a proper slice definition, two subsequent excitations with slice select gradients of opposite sign are applied (Fig. 1), and their MR signals are added. The half-sinc pulse is inherently self-refocused [2] and is applied also during the falling edge of the slice selection gradient. A fast TwiRL read-out is used [3]. The half-sinc excitation approach suffers from sensitivities to gradient and main field imperfections [4]. The phase error caused by B0 eddy currents and gradient waveform distortions can be expressed as ϕ (r,t) = ϕB0 (t) + rk(t). B0 eddy currents cause a spatially invariant phase shift, whereas short-time-constant eddy currents result in a delay of the gradient waveform. Results: Experiments were performed on a Philips 3T whole body scanner using a local T/R coil with a switching time of 60 µs. To characterize the effects of B0 eddy currents, the 1D slice profile was measured in a preparation scan as a function of the phase offset between the signals from the two excitations (Fig. 2). It was observed that there exists an optimum phase offset, which was found to be system and orientation dependent and, thus, must be determined before each scan. Using the UTE imaging sequence, in vivo images of a transverse

S232

Imaging: Sequences and Techniques

slice through the foot were obtained (Fig. 3: TE = 1000 µs, 280 µs, 160 µs). Conclusion: Enhanced signal from the bone structures in the foot are observed with ultrashort echo times. This signal enhancement is believed to be caused by the short T2 components in the bone tissue. References: 1. Gatehouse P et al. [2003], Cl. Rad. 58:1-19 2. Pauly J et al. [1989], Proc. SMRM, 28 3. Nielsen H et al. [1999], MRM, 41:591-599 4. Wansapura J et al. [2001], MRM, 46:985-992

439 The use of a T2 weighted image acquisition as saturation recovery preparation for a T1 weighted image acquisition: a new MR acquisition approach for ultrafast sequential multicontrast MRI screening H. T. C. Bosmans1, F. De Keyzer2, J. Michiels3, K. Scheffler4, M. Bock5, G. Marchal1; 1Radiology, University Hospital Gasthuisberg, Leuven, BELGIUM, 2Radiology and Medical Imaging Computing, University Hospital Gasthuisberg, Leuven, BELGIUM, 3Siemens, Medical Systems, Brussel, BELGIUM, 4Medical Radiology, University Hospital Basel, Basel, SWITZERLAND, 5AG Interventionelle Verfahren, DKFZ, Heidelberg, GERMANY. Purpose: To validate and optimize a new ultrafast spin echo based MRI technique that provides for each sequential slice a T2, heavily T2 and T1 weighted image Material and Methods: The new MR sequence acquires 3 series of spin echo trains. After a 90° RF excitation pulse, a first echo train is acquired for a T2 weighted single shot TSE image, followed by a second echo train to generate a heavily T2 weighted image. These pulses saturate the tissue in the particular slice. After a ‘saturation recovery' delay of 400ms, a new 90° RF pulse is given, followed by another complete TSE read-out. Theory predicts this image to be T1 weighted. The sequence was implemented on a 1.5T system (Sonata, Siemens, Germany). Image contrasts were validated in phantoms and in selected patients. The effects of scan percentage and half Fourier techniques were evaluated. Results: The implementation of the sequence required dedicated pulse sequence programming. Phantom and abdominal MRI examinations confirmed the T2 and heavily T2 weighted image contrast of the first images, and the T1 weighted contrast of the third image. Image quality of the T1 weighted image was acceptable. Three images with FOV 225 x 320 mm, matrix 112x256, 80% scan percentage and half Fourier acquisition (i.e. only 192 RF pulses for 3 images), were typically obtained in a total time of 1.3s. Image sharpness of the T1 weighted image improved significantly if half Fourier was not activated. The associated SAR limitation led to the use of refocusing flip angles of about 110°. No heath sensation was noticed. A whole upper abdomen can be scanned in less than 1 minute. Breath hold was not necessary. Partial volume effects were absent: the same anatomical slice is visualized with 3 image contrasts. Discussion: Image quality of the T2 weighted images is well known from daily routine. The quality of the T1 weighted image is probably sufficient for screening purposes, for MRI in uncooperative patients and for perfusion studies. Image sharpness may be further improved by parallel imaging and the use of shorter echo spacing. Keyhole imaging could further reduce the number of RF pulses. These options are being evaluated. Conclusion: First validation studies of a new sequential scanning technique, that uses the T2 weighted image to prepare the T1 weighted contrast, underlined its potential for rapid scanning of a large anatomical region and combined T2 and T1 weighted perfusion studies.

Imaging: Sequences and Techniques 440 Systematic sequence performance evaluation: Fast Gradient Echo Sequences (TFE) for contrast-agent enhanced renal perfusion protocols R. L. Perrin, MSc, M. K. Ivancevic, MSc, J. Vallée, MD,PhD; Radiology, University Hospital of Geneva, Geneva 14, SWITZERLAND. Purpose/Introduction: MR techniques has been shown to be a powerful tool with use of Gadolinium-DTPA (Gd-DTPA) contrast medium for renal perfusion imaging [1],[2]. Two parameters are important with respect to the signal intensity (SI) versus [GdDTPA] curve: dynamic range (concentration range over which the SI increases) and sensitivity i.e. slope of the curve before saturation. This study carried out a systematic study of the influence of sequence parameters on the dynamic range and sensitivity of new fast TFE (Turbo Field Echo) sequences. Subjects and Methods: Tubes of Gd-DTPA at a range of concentrations ([Gd-DTPA]=0-50mM) were used to produce the SI vs. [Gd-DTPA] curves. Three magnetisation-prepared TFE sequences were studied on a Philips 1.5T Intera MR system while varying sequence parameters. Sequences: TFE (viz.FAST), T1W-TFE (viz.RF-spoiled FAST), and balancedTFE (viz.true-FISP). Parameters: flip-angle (FA), echo-time (TE), inversion-time (TI), preparation pulse, partial-echo and shared-preparation for spatial slices. Sequence parameters (except when varying a specific variable): TR=3.3ms (shared/non-shared 2.8ms), TE=2.1ms (shared/nonshared 0.93ms), FA=90°,*TI=28ms, pre-pulse FA=180°, matrix=192x256 (shared/non-shared 83x128), FOV=46cm, 10 dynamic scans. Any variation in receiver gain was removed by using the Philips scaling factor, ss (ss~1/gain). SI was plotted as a function of [GdDTPA] for all sequences to quantify sensitivity and dynamic range. * Time between pre-pulse and 1st-excitation pulse Results: For T1-TFE the contrast dynamic range increased with FA (10°-90°-see fig.1a). For b-TFE (fig.1b) the sensitivity increased with FA. No-TFE gave similar dynamic range to b-TFE with a lower sensitivity. The TE (1.5-5ms) and partial echo had no effect on the sensitivity of the b-TFE and no-TFE sequences. However, the sensitivity of the T1-TFE sequence could be varied with TE (increase TE, increase sensitivity). There was no notable difference in sensitivity and dynamic range with TI (668-1168ms) or pre-pulse FA (90/180°) for each sequence. With the shared option the sensitivity decreased by 60% between 1st and 4th slices (whereas nonshared yielded no difference in sensitivity).

Discussion/Conclusion: This study revealed important differences between the sequences. For quantitative analysis, where contrast dynamic range is important, the magnetisation-prepared T1-TFE sequence with high FA and short TE is optimal. For qualitative analysis, where sensitivity is important, the magnetisation-prepared

S233

b-TFE sequence with a high FA and short TE, is optimal, as opposed to the no-TFE which was favoured previously. References: 1. Laissy JP et al.(1994) Magn Reson Imaging ;12(3):413-9 2. Vallee JP, et al.(2000) Eur Radiol ;10(8):1245-52.

441 Reducing SAR while maintaining optimal image quality J. Beusen, T. Cox, K. Munter; MR-applications, Philips Medical Systems, Best, NETHERLANDS. Introduction: This project investigates the possibility of decreasing the Specific Absorption Rate (SAR) in T2W-TSE sequences, while maintaining optimal image quality. The maximum SAR-level of 4.0 W/kg* can be reached on 1.5T systems with a high gradient performance (high amplitude, high slew rate) in combination with high RF transmit fields. The problem occurs mainly in TSE sequences with high turbo factors due to the large amount of RF power deposition. High SAR-levels can increase patient temperature, this may cause discomfort. *IEC 60601-2-33 guidelines Methods: In this investigation, the SAR was decreased by adjusting the B1-amplitude and/or the refocusing flip angle. The following series were scanned on a 1.5T system with a high gradient performance, a maximum RF transmit field of 27 µT and a fixed TE: • A variable B1-amplitude, ranging from 11 µT to 27 µT. • Variable refocusing flip angles, ranging from 90° to 160°. Three different regions of the body were scanned on volunteers; the pelvis, the abdomen and the knee. The SNR, CNR and sharpness of the images were scored subjectively by 29 Radiologists. Results: The Radiologists’ evaluations were processed, these results are displayed below. The score is displayed versus the B1-amplitude (top) and the refocusing flip angle (bottom). The length of the bars indicates the averaged scores of the SNR, CNR and sharpness. The value above each bar is the corresponding SAR.

Conclusions and discussion points: From the processed data the following can be concluded: • The SAR can be reduced by using a lower B1-amplitude with no significant loss of image quality; amplitudes of 15 µT and 17 µT even showed an improvement of the SNR, CNR and sharpness. • It is possible to reduce the SAR by means of altering the refocusing flip angle when using an initial refocusing flip angle of 135°. No significant loss of SNR, CNR and sharpness occurs here. When using an initial refocusing flip angle of 180°, loss of image quality does occur.

S234 •

• •

•

Imaging: Sequences and Techniques

The 135°/120° scan was evaluated as the best image of the non-default scans. Discussion points for this investigation are: Could a lower B1-amplitude be a solution for non-TSE sequences? Is simultaneous adjustment of the B1-amplitude and the refocusing flip angle a good option (especially when using an initial refocusing flip angle of 135°)? Are the 1.5T solutions applicable to 3T systems?

442 Application of simultaneous fat and water signal suppression for 3D visualization of intracranial pathological forms N. V. Anisimov1, L. V. Gubskii2, Y. A. Pirogov1; 1Center of Magnetic Tomography and Spectroscopy, Moscow State University, Moscow, RUSSIAN FEDERATION, 2Chair of Neurology and Neuroradiology, Russian State Medical University, Moscow, RUSSIAN FEDERATION. Purpose: For enhancement of visualization and 3D reconstruction of pathological forms we use MR images obtained with simultaneous water and fat signals suppression (SWFSS). In this case one can more distinctly reveal pathological forms that are hidden under powerful signals of fat and water. To realize SWFSS, we modified the “inversion-recovery” pulse sequence. As tissue contrast picture is maximally refined the graphical data processing and construction of 3D images are significantly simplified. Subjects and Methods: For SWFSS we combined two inversionrecovery pulse sequences – FLAIR and STIR to read MR signal in a moment when longitudinal magnetization of both water and fat passed through zero [1]. We realized pulse sequence: 1800-TIw1800-TIf-900-FID, where TIf =80 ms, TIw =1300 ms. As pathological forms possessed intermediate longitudinal relaxation times they gave strong enough signal. Because strong background signals were suppressed, the sensitivity of receiver was increased. Drawing of 3D images was carried out by standard software of MR scanner TOMIKON S50 (Bruker). We used axial MR images obtained by SWFSS. Standard software gave possibilities to interpolate slice data into 3D matrix and construct 3D image with different variants of space orientation, illumination and animation effects. Results: By SWFSS we obtained the simplest distribution of contrast at MR image. It in turn simplified the graphical treatment of images for the MIP reconstruction, the volume rendering and morphometrical measurements. This method was used at 38 patients for head, abdomen, prostate and hip joint MRI investigations. It was particularly attractive for intracranial structure investigation, for example of dynamics of subdural hematomas (Fig.1), intracranial tumors and multiple sclerosis.

Figure 1: The simultaneous water and fat signals suppression at subdural hematomas. From left to right – MR image (axial projection), MIP-reconstruction, 3D rendering. Upper set – the first study after intracranial injury, middle set – the second study after 9 weeks, lower set – the third study after 20 weeks. Discussion: If pathological forms have relaxation times other than fat and water, it is useful to maximally simplify the picture of MRI tissue contrast by simultaneous water and fat signals suppression. In this case the visualization and 3D graphical data processing of pathological forms are significantly simplified. Reference: 1. Pirogov YuA, Anisimov NV, Gubskii LV [2002] Proceedings of SPIE, 4681:612-616.

443 Correlation between relaxometry determined and tMIX estimated T1-T2 values in a phantom: validation study I. Van Breuseghem1, L. Vander Elst2, H. Bosmans1; 1Radiology, University Hospitals Leuven, Leuven, BELGIUM, 2Organic Chemistry - NMR Labaratory, University of Mons - Hainaut, Mons, BELGIUM. Purpose/Introduction: The turbo-mixed T1 and T2 quantification sequence (tMIX) had been used in our department for T2 maps of cartilage on a clinical 1.5T system. In this study, we evaluated the correlation of T1 and T2 values determined with this tMIX sequence and with an accepted spectroscopic method. Subjects and Methods: A phantom comprising 7 tubes with different gadolinium-DTPA/water solutions was made (table 1). Table 1 Tube Number 1 2 3 4 5 6 7

Dilution 1/20 1/40 1/50 1/100 1/250 1/500 1/1000

Imaging: Sequences and Techniques The T1 and T2 relaxation time of these solutions (heated to 37°C) were determined with relaxometry measurements on a Bruker minispec Mq 60 (Bruker Analytik, Rheinsteffen, Germany). The tMIX sequence was run on a 1.5T MR imaging system (Gyroscan, Philips Medical Systems, Best, The Netherlands). The sequence consists of interleaved spin-echo (SE) and inversionrecovery (IR) acquisitions. The SE counterpart has TR 1945 msec and echo times 7.4 and 70 msec; the IR part has an IR-delay of 400ms, TR 1545 msec and echo times 7.4 and 70 msec. T1 and T2 relaxation times of the same solutions, heated to 37°C, were estimated with this technique and displayed using the built-in calculation algorithm. The correlation between spectroscopy determined and tMIX estimated T1 and T2 values was calculated with Pearson correlation coefficient. Relative differences between both values were calculated. Results: The graphs show the spectroscopically determined versus tMIX-estimated values for T1 and T2. A high correlation between these values is shown for T1 (r=.99) as well as for T2 (r=.99). One can note a systematic error in tubes 1-4 and a low relative difference in the highly diluted tubes (less than 20%) for T1; for T2 relative differences are low (less than 14%) except for tube 7 (with a relative difference of 45%).

Discussion/Conclusion: Errors need to be considered when using the tMIX method for T1 and T2 determination: calculation errors are due to the 2 point curve fitting, whereas it is generally fit using 3 points. Stimulated echo and magnetization transfer contribution

S235

influence the measurement, and diffusion weighting is introduced by the long inter-echo interval. Obtained values correspond well in the expected cartilage relaxation times, i.e. for T2 between 20 and 100 msec. For T1 in the interval 150ms to 400ms, a small systematic value underestimation is noted; when corrected (by multiplying with factor 1.25), relative differences less than 6% are calculated. In conclusion, tMIX calculation of T1 and T2 relaxation times is a valuable technique especially in the range of cartilage relaxation times.

444 Establishing important clinical contrasts and high resolutions in hyperecho-TSE's M. Weigel, J. Hennig; Section of Medical Physics, Department of Diagnostic Radiology, University Hospital of Freiburg, Freiburg, GERMANY. Purpose: Turbo-spinecho-sequences (TSE) are the workhorse in diagnostic MRI due to their clear T1- and/or T2-weighted contrast behavior and insensitivity to susceptibility effects. Recently, novel acquisition techniques for TSE-sequences such as hyperechoes[1] and TRAPS[2] have been proposed to reduce RF power deposition while maintaining SNR and a similar contrast (hyperTSE-sequences). This contribution shows that hyperTSE-sequences can produce high quality IR-weighted in addition to T2-weighted images, but with significantly reduced SAR. We also demonstrate that these techniques can produce high resolution images without any loss of quality at 3T. Subjects and Methods: All experiments were performed on a 3.0T whole-body imaging system (Siemens Trio, Erlangen, Germany). A common TSE-sequence (MTX=512x408, in-plane resolution<=(0.4mm)2, ETL=27, ESP=6.6ms, slth=2-3mm) was employed, in which the “TRAPS”-mechanism was implemented. Optimized sinusoidal flip angle variations were used such, that the flip angles were ramped down from 180° to 60° in the outer parts of k-space acquisition. Contrast specific parameters were: a) T2weighting: TE/TR=101/5000ms, AVE=4; b) IR-imaging: TE/TR=10/7000ms, TI=800ms, AVE=3. Results: Fig.1 shows three of 15 possible T2-weighted images acquired with the described hyperTSE-sequence. Each image displays a strong T2-weighting at a high resolution with a high SNR for excellent neurological diagnosis. In Fig.2 typical IR-contrast images are presented (five of 13 possible slices). Notice the very good contrast to noise ratio between cerebral WM, GM and CSF. The contrast is identical to a conventional IR-TSE180° sequence. Acquisition times for both types of images were about 5min. All employed sequences showed a relative SARrel reduced to about 33% for T2-weighting and approximately 45% for IR-weighting compared to conventional TSE180°. Discussion and Conclusion: It was demonstrated that hyperTSEsequences are capable of acquiring T2- and IR-weighted images with a high resolution and no degradation in image quality. SAR deposition can be reduced by as much as 70%. This reserve in SAR was used for achieving longer echo train lengths (leading to shorter acquisition times) and a larger amount of slices. For example, a standard T2-weighted TSE180° with identical acquisition parameters at 3T would have allowed only five slices instead of 15. 1. Hennig J, Scheffler K, MRM 46:6-12 2. Hennig J, Weigel M, Scheffler K, MRM 49:527-35

S236

Imaging: Processing and Quantification Only 13 sets of reconstruction parameters were calculated, covering shifts from -3 ∆ky to 0 ∆ky in steps of 0.25 (Fig.1,right). The reconstruction parameters vary slowly over the fractional harmonic values, so that all intermediate shifts necessary for the reconstruction can be easily interpolated. Results: In order to demonstrate the feasibility of our approach, results from a VD-acquisition are shown in Fig.2. The DFT of a VDdata set with 128x256 matrix size is shown in Fig.2a. Aliasing artifacts are clearly visible as a result of the undersampled nonuniform FOV. Application of our approach to the same data set leads to significantly reduced foldover artifacts and reduced blurring, see Fig.2b. Discussion: In this work we have demonstrated that it is feasible to use interpolated reconstruction parameters for non-Cartesian parallel imaging. The flexible implementation of our approach allows one to use other regenerative k-space based parallel imaging methods, such as AUTO-SMASH(5) or VD-AUTO-SMASH(6) to determine the reconstruction parameters. This technique could also easliy be extended to 2D non-Cartesian trajectories. References: 1. Yeh,ISMRM2002#2390 2. Yeh,ISMRM2001#1796 3. Pruessmann,MRM2001;46:638-651. 4. Kyriakos,MRM2000;44:301-308. 5. Jakob,MAGMA1998;7:42-54. 6. Heidemann,MRM2001;45:1066-1074.

445 Fast non-cartesian parallel image reconstruction R. M. Heidemann, M. A. Griswold, M. Müller, M. Blaimer, F. Breuer, P. M. Jakob; Department of Biophysics, EP 5, University of Wuerzburg, Wuerzburg, GERMANY. Introduction: Non-Cartesian sampling presents unique problems for parallel imaging. In general, large linear systems are solved to reconstruct images from undersampled data (1). An alternative was proposed by Yeh(2), which uses non-integer shifts for a SMASHlike reconstruction. This method has been extended so that more than 1 point is used in each reconstruction, however this results in drastic increases in reconstruction time, beyond that of conjugategradient methods (3). In this work variable density (VD) trajectories are used, similar to the VD-sampling proposed in(4). The distance between adjacent lines in this acquisition grows continously from a minimal to a maximal value. In a first step, lines at intermediate positions between measured lines are reconstructed using a SMASH-like procedure with fractional harmonics. Instead of using a separate inversion for each fractional harmonic the reconstruction parameters for a small number fractional harmonics are calculated allowing the weights for intermediate harmonics to be determined by interpolation, resulting in increased reconstruction speed. A direct Fourier Transformation (DFT) is finally applied to the data to obtain the image. Methods: The distance between adjacent lines in the VD-sampling scheme varies continiously from 0.5 to 3.0 ∆ky. A cartesian acquisition from –128 to +127 ∆ky, requires 256 phase encoding steps. The same range of ky can be covered with 128 VD-lines on a undersampled non-Cartesian grid.

446 Evaluation of the new i-PAT head coil and partially parallel acquisition technique (GRAPPA) in cerebral imaging: A phantom study J. Tintera1, J. Gawehn2, T. Bauermann2, P. Stoeter2; 1Department of Radiology, IKEM, Prague, CZECH REPUBLIC, 2Institute of Neuroradiology, University Clinic, Mainz, GERMANY. Introduction: A newly-designed 8-channel head coil array allowing for parallel acquisition of independently reconstructed images (1) was tested for its applicability in neuroradiology. Methods: The new i-PAT head coil array consists out of 8 surface coils and was implemented in a 1.5 T Sonata scanner (Siemens,

Imaging: Processing and Quantification Erlangen). Standard and GRAPPA mode were used to test the image quality and results were also compared with a standard head coil. 1) Image homogeneity was tested using a 3D FLASH sequence (TR=15ms, TE=5ms, flip angle (FA)=15°, matrix=256x256, FOV=256mm, slice thickness=1mm, 176 slices).Three measurements were compared: a) using standard head coil and using i-PAT coil, b) without and c) with normalization filter. 2) Image quality expressed over artifact power was measured with a) 3D FLASH (TR=15ms, TE=3ms, flip angle=15°, bandwidth=260 Hz/pixel, matrix=256x256, FOV=230mm, slice thickness=5 mm, 32 slices) and b) GE EPI sequence (TR=6s, TE=59ms, FA=90°, bandwidth=1300 Hz/pixel, matrix=128x128, FOV=230mm, slice thickness=4mm). Twelve measurements with various acceleration factors (AF=2,3,4) and number of reference lines (RL=8,16,32,64) were performed with FLASH 3D and 4 with EPI sequence: a) AF=2, RL=64, b) AF=2, RL=24, c) AF=3, RL=64 and d) without GRAPPA. Results: 1. Maps of relative signal homogeneity created from 3D data sets acquired by the standard head coil and the i-PAT coil without and with normalization filter were compared. Signal homogeneity was best preserved in the standard head coil compared to the i-PAT coil but improved with the normalization filter. 2a. Combination of various AF and RL can result up to 70% of the acquisition shortening but this leads also to the increase of artifact power and image degradation (fig.1), moreover there is no linear relation between these parameters.

Fig.1.Subtraction of 3D FLASH GRAPPA and standard image 2b. Susceptibility artifacts on EPI images were clearly reduced only with the combination of AF=2 and RL=64 but other parameters could lead to even larger area of signal drop out (fig.2). Artificial signal variations calculated as artifact power grew from 2.27 (2a) to 2.66 (2b) and 4.48 (2c).

Fig.2.Subtraction of GE EPI GRAPPA and images without PPA.

S237

Discussion: Confirming previous examinations, we also found a decrease in SNR and an increase in artifact power with enlargement of the AF. As the latter also depends on the RL, a combination of a higher AF with a higher RL is more effective in artifact reduction within a similar acquisition time. 1. Griswold MA et al. [2002] Magn.Reson.Med 47:1202-1210

Imaging: Processing and Quantification 447 Mapping human cerebral water compartments based on simultaneously acquired T1 and T2 data P. Vermathen1, J. Pietz2, C. Boesch1, R. Kreis1; 1Dept. Clinical Research, AMSM, University and Inselspital, Bern, SWITZERLAND, 2Dept. of Pediatric Neurology, University of Heidelberg, Heidelberg, GERMANY. Introduction: Separate cerebral water compartments lead to multiple relaxation time components, mostly measured as T2-dispersion. Relatively little is known about the water component in myelin layers that has short T1 and T2 relaxation times. Relaxometry based on an inversion recovery (IR) multiple-echo sequence was used to characterize and map four brain compartments: tissue water in gray (GM) and white matter (WM), CSF, and water in myelin layers. The method was tested on volunteers and two patients with phenylketonuria (PKU) with known WM abnormalities, tentatively attributed to dysmyelination. Method: An IR fast-spin-echo sequence was modified to yield 32 images, each acquired with an individual TE (11ms spacing). The sequence was repeated with 5 different recovery times TI (102010ms). Thus, for each pixel 180 intensity values with different combinations of TE and TI were obtained. Imaging time for one slice (matrix: 512 ×16) was ~5min. Five volunteers and two PKU patients were measured. In control subjects, two transverse slices were acquired, one including and one slightly above the ventricles. In patients, one slice was acquired including the known T2-hyperintensities. One patient was measured twice within 24h. The data were fitted to yield M0, T1, and T2 of all four compartments in each pixel. For the presented results, T1, T2 were fixed for WM and GM and constrained for the myelin component and CSF. Results: Figure 1 shows images of the different compartments for two slices. Images of WM, GM and CSF are well separated. The images of the myelin component show intense signals only in areas with high WM content. Overall, a content of 14.7±2.5% myelin component (% of WM) was found for control subjects (Figure 2). The two different slice positions showed similar results, while PKU subjects had lower myelin contributions of 10.1% and 8.3%, respectively. The repeated scan on one patient yielded almost identical results (10.1% vs. 10.2%). Discussion: Fitting of simultaneously acquired T1- and T2-data yielded consistent brain components. The method appears robust (low intrasubject variations for the myelin water content from multiple voxels) and suitable to characterize focal or diffuse cerebral abnormalities in pathology, even though component sizes or relaxation times may be systematically influenced by the enforced prior knowledge constraints. The average myelin component of 14.7% (relative to WM) matches literature values. The preliminary results on PKU patients confirm the concept of dysmyelination as cause of the known MRI abnormalities. Acknowledgment: Swiss National Foundation (31-059082)

S238

Imaging: Processing and Quantification Results: In order to facilitate comparisons the patients were separated into two groups dependent on response at TP2 – those who showed less than 50% decrease in tumour volume and those who showed greater than 50% reduction in tumour volume. Significant (p<0.05) or borderline significant (0.05
448 Textural analysis as a predictor of breast tumour response to neoadjuvant chemotherapy P. Gibbs, M. Lowry, L. W. Turnbull; MRI Centre, Hull Royal Infirmary, Hull, UNITED KINGDOM. Purpose/Introduction: Neoadjuvant chemotherapy is routinely used to improve outcome in patients presenting with inoperable breast cancer. Assessment of tumour response is crucial to patient management and is conventionally assessed by clinical examination and X-ray mammography. However, these methods have limitations, particularly in the presence of dense fibroglandular tissue. Since a poor response to treatment will prompt a change to second line chemotherapy a more reliable indicator of early response is required. MRI has shown some promise in this area [1]. This work aims to determine whether textural analysis of post-contrast images has a role to play in the assessment of tumour response to chemotherapy. Subjects and Methods: Forty-one women (25-76 years old; mean = 50 years) with inoperable primary breast cancer were treated by standard dose protocols using intravenous 5-fluororacil, epirubicin and cyclophosphamide. MRI examinations were performed prior to treatment (TP0), between treatments 2 & 3 (TP2), and subsequent to treatment 4 (TPF). Along with dynamic contrast enhanced imaging, spectroscopic imaging, and diffusion weighted imaging, highresolution (voxel volume from 0.46 mm3 to 2.54 mm3) post-contrast images were obtained. Tumour volumes were manually segmented and ranged in size from 3.3 cm3 to 234.0 cm3 (1536 to 125785 pixels). After histogram equalisation and decimation to 32 grey levels, co-occurrence matrices were calculated. Textural analysis was implemented using the spatial grey level dependence method [2].

449 Texture analysis of calf muscle MR images A. Škoch, P. Fendrych, E. Rolencová, P. Harantová, M. Dezortová, M. Hájek; MR-Unit, Dept. Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague, CZECH REPUBLIC. Purpose: To perform texture analysis (TA) of MR images of healthy and affected calf muscles. Subjects and Methods: In total 117 informed subjects (20 healthy controls Con1, 7 healthy children of hypertonic parents ConH, 14 obese women ObeP, 5 patients with diabetes mellitus DiaP and 71 subjects with muscle malfunction of various cause DifP) underwent MR imaging of the calf on Siemens 1.5T MR scanner. Images were acquired with constant protocol for all patients and controls using T1W SE sequence (TR/TE=500/12ms, NA=1, FOV=200x200, matrix 256x256, thickness 6mm) and T2 TSE sequence (TR/TE=5400/99ms, TF=11, NA=1, FOV=200x200, matrix 198x256, thickness 6mm). Texture features of calf images were computed by program MaZda [1] after manual selection of ROIs within 3 different calf muscle types (tibialis anterior, soleus, gastrocnemius). Radiological assessment reflecting the stage of

Imaging: Processing and Quantification muscle pathology was done by senior radiologists and images of DifP were divided in this manner into 4 groups from DifP1 (subjects with muscles assessed as normal) to DifP4 (subjects with most severe muscle fat degeneration). This categorization of the subjects was compared with the results of principal component analysis (PCA) of texture features. Results: We selected 33 texture features with maximal difference between control group and randomly selected patients with severe muscle malfunction. PCA of 33 features (see Fig.1) showed correlation between some of them and provided possibility to reduce their amount to only 6 features which were used for classification of subjects by PCA. With the help of PCA we were not able to differentiate the Con1, ObeP and ConH groups, while separation of the groups Con1 and DiaP was quite good. In case of DifP we compared the classification from PCA with the assignment of patients and controls by radiologists. We constructed borders between clusters of subjects categorized by radiologists into 4 stages of muscle pathology (see Fig.2). Using this method the error of assignment of the patient into the proper cluster was found 19%. Main contribution to this error is unclear separation of DifP1 and DifP2 groups. We consider the results of automatic classification by TA in good agreement with the categorization done by the radiologist. Conclusion: The analysis of texture features can be used for the quantitative description of changes in muscles and could serve as a tool which can help radiologists to distinguish healthy and diseased tissue. This study was supported by grant CEZ:L17/98:00023001. Reference: 1. http://www.eletel.p.lodz.pl/cost/progr_mazda.html

S239

450 Removal of g-factor enhanced noise in Partially Parallel Imaging using information entropy D. J. Larkman, J. V. Hajnal; Imaging Sciences Department, Imperial College, London, UNITED KINGDOM. Introduction: Partially Parallel Imaging (PPI) uses array coil sensitivity information to allow image reconstruction from reduced data sets. The signal-to-noise ratio is degraded by a factor g that depends on coil geometry [1]. For fast imaging, the acceleration factor may be limited by unacceptable increases in the g-factor. We have investigated a method to remove the g-factor noise contribution. Method: Using an array of m coils with a field of view decreased by a factor m, we reconstruct using the SENSE method to separate signals from families of m pixels that are aliased together. Each family of pixels is processed separately and the resulting image suffers from spatially dependent noise amplification (the g-factor effect). If S is the aliased pixel data from the coils, n is the explicit noise from the coils, C(x) a matrix of coil sensitivity factors and X the reconstructed pixel values, then X=C-1(S + n). Noise in the final image is given by the spatially varying product C-1n, which we wish to remove. Although the elements of n are unknown, each noise component effects m pixels in a deterministic way. Image entropy [2] is a measure of total information content defined as E = -Σpi log(pi), where pi is the probability of a pixel having intensity i. The pi can be determined directly from a histogram of X. The presence of C-1n increases the entropy. Minimising E as a function of estimates of n allows the amplified noise to be determined. This concept was tested using simulated test images of a phantom, the Shepp-Logan brain phantom and a T1 weighted image from the Brainweb phantom [3]. Simplex optimisation was performed one image line at a time, with Entropy calculated for the complete image. Results: Figure 1 shows a typical result for factor 4 speed up with

S240

Imaging: Processing and Quantification

4 coils and 0.25% noise added to each channel. In all cases tested, there was a substantial reduction in signal fluctuations. Discussion: Image entropy provides a basis for removing fluctuations introduced by PPI reconstruction. It provides an unbiased image wide metric for removing g-factor effects, but may be vulnerable to local errors. Further tests will be required to determine how universal this approach is. References: 1. Pruessmann KP et al. MRM 1999;42:952-962 2. Shannon CE. IEEE 1998;86:447-457 3. http://www.bic.mni.mcgill.ca/brainweb/

Firstly, the effect of the noise is reduced by smoothing the data. Secondly the square of this smoothened signal is calculated. Thirdly, smoothing and scaling reduce the effect of the mixed term of equation 2. Then the power of the smoothened noise is considered as the square of rms noise ( σ2). A reference noise σ'2 is calculated and subtracted (equation 3). Finally the hat window is taken as the square root of equation 3 . Results and application. A K-Space obtained on a Bruker-Avance 200 has been treated. The K-space of a shallot was recorded with a SNR(image)=1.7 (figure 2A). The corresponding hat window calculated (figure 2B) is close to the envelope function of the original K-space. Three processing methods were applied using the hat window: 1. The nst-root of the matched filter (equation 5) (figures 2 C-D). 2. A time transfer function (3) ϕ(t) (equation 6), (figure 2E) . 3. TRAF window function(4) (equation 7, figure 2F).

Figure 1: Left half-Conventional SENSE reconstruction; right halfEntropy minimisation reduces noise in areas where there is g-factor noise amplification.

451 Matched filter for NMR signal H. J. Tellez, A. Roch, Y. Gossuin, R. N. Muller; Dpt Organic Chemistry/NMR Lab, University of Mons-Hainaut, Mons, BELGIUM. Introduction: The fundamental limitation of MRI is signal-tonoise ratio (SNR). To improve the final quality of the images windowing methods are usually applied (1). The matched filter, which is the function matching the envelope of the signal is very important to maximize the SNR (2). In this work, a numerical method called hat window (hw(t)), is proposed to obtain an approximation of this envelope function. Methodology: A noisy signal can be represented by the sum of a NMR signal m(t) and a noise signal n(t). Conclusion: Original numerical method is proposed to calculate the matched filter for the NMR signal. This new hat window is a fast procedure which improves the SNR without significant loss of resolution. References: 1. Lindon JC, Ferrige AG. [1980]Progress in NMR Spectroscopy 14:27-66. 2. Ernst RR, Anderson WA[1966]The review scientific instruments 37(1):93-102 3. Proakis JG, Manolakis DG [1996] Digital signal processing. Prentice-Hall International Edition. 4. Traficante DD,Rajabzadeh M. [2000]CMR, 12(2):86-101

Imaging: Processing and Quantification

S241

452 Device for validation of biomechanical models in MRI-guided neurosurgery D. F. Kacher1, A. Balakrishnan2, C. Kemper3, A. Slocum2, K. Chinzei1, F. Talos1, F. A. Jolesz1, S. K. Warfield1; 1Radiology, Brigham and Women's Hospital, Boston, MA, 2Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, 3Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA. Introduction: Intra-operative MRI-guided removal of brain tumors typically utilizes low- to mid-field scanners, characterized by limited image quality [1]. Moreover, time dedicated to image acquisition is curtailed by the desire to prevent extended procedure time. Images acquired prior to surgery are not subject to these limitations and are useful for guiding surgery. However, deformations in the brain shape or “brain shift”, occurring over the course of surgery quickly reduces the accuracy of pre-operative images. An intra-operative non-rigid registration algorithm and biomechanical model have been developed to elastically match pre-operative images to intra-operative images during neurosurgery, to enable continued guidance with the pre-operative images [2]. We have developed a device, the “smart retractor”, to validate such models, without introducing additional invasiveness to the procedure. Methods: A cube of gelatin with imbedded glass beads was compressed in a test jig and MR images were acquired with no load and five levels of incremental compression. The surface of the gelatin was coated with oil to allow lateral movement. A smart retractor was placed between the compression piece and the gelatin (Fig. 1).

The smart retractor, described elsewhere [3], is a strip of semi-rigid copper with four force sensors (Flexiforce®, Tekscan Inc., Boston, Ma) (Fig. 2). The model assumes a linear elastic, isotropic material, and uses material property values based on in-vivo experiments with porcine brain [4]. Surface displacement is derived from the images, material properties are assumed, and force and subsurface displacement are calculated by the model.

Results: All six volumes were segmented and the first volume elastically matched to the subsequent five time points using the model. Three representative volumes are superimposed in Fig 3: ‘No load’ (red), Load 5 (blue), No load elastically matched to Load 5 (yellow). The grayscale images showing the glass beads are below.

Table 1 shows the force-displacement data imaging experiment.

No load Load 1 Load 2 Load 3 Load 4 Load 5

SI (mm) 55 59 60 64 63 71

RL (mm) 57 58 60 64 65 70

AP (mm) 38 35 32 31 29 25

Sensor 1 (kg) 0.12 0.13 0.13 0.14 0.14 0.15

Sensor 2 (kg) 0.12 0.13 0.13 0.14 0.14 0.15

Sensor 3 (kg) 0.08 0.08 0.07 0.07 0.06 0.05

Conclusion: It remains to compare the force data calculated by the model to measured force data. The location of the glass beads can be used to evaluate the accuracy of the model deformation for deeper structures. The device has been developed to a point where this comparison is now possible. References: 1. Black et al.,. Neurosurgery, 41:831-842, 1997 (1998). 2. Ferrant M. et.al.,. IEEE Trans. Med. Imag., 20:1384-1397 (2001). 3. Balakrishnan A. et al., Proc. ASME (2003). 4. Miller, K. et al.,. J Biomech, 33:1369-1376 (2000). Acknowledgements: NAC grant RR013218-06.

453 MRI tumor segmentation using wavelet transform A. Lopez1, Y. Brahami1, M. Kelarestaghi1, A. Ouahabi2, S. Benderbous1; 1Inserm U316, UFR Sciences Pharmaceutiques, Tours, FRANCE, 2Laboratoire traitement du signal, Ecole Polytechnique de Tours, Tours, FRANCE. Purpose/Introduction: Image segmentation plays a crucial role in many medical imaging applications, by automating or facilitating the delineation of anatomical structures. We describe simple and effective technique for images denoising and tumor delineation based on the multiscale edge representation of the images. The method is based on the maxima chaining through the scales of the decomposition helping to select the largest discontinuities of the signal in order to segment it. A comparison of different 2D-wavelet transform applied on real MR images is performed. Subjects and Methods: Experimental model : At the age of mammary gland maturation, 15 Sprague-Dawley rats received a single subcutaneous injection of

S242

Imaging: Processing and Quantification

25 mg/kg of N-methyl-nitrosourea. Time to tumor appearance, growth and location of the tumors within the mammary gland chains were monitored by weekly palpation. The rats were imaged when tumor size reached 1 to 2 cm using a bruker imager at 2.35T. T1, T2 and proton-density weighting (100 * 100 µm2resolution) were acquired before and after caudal intravenous injection of GdDOTA (0.2 mmol Gd/kg, n=6) or after iron oxide nanoparticles (80 µmol Fe/kg, n=9). Images processing : After background suppression using a Wiener filter or suppression of low correlated wavelet coefficients, images are decomposed into two spaces (approximate and residual matrix) by the wavelet transform. Edges which do not belong to the principal edge of the heterogeneous tumour are cancelled. Results: Contours of the tumor are not sufficiently defined on small scales (see figure b). Edges are visually well restored from regularized multiscale edges. When increasing the wavelet scale, contours of smaller tumor-nodules appeared more cleary (c,d).

As the tumor is placed near the abdominal area (figure a), respiratory movement induce main noise - background. Concerning the suppression of the background noise, the Wiener filter appears more powerful than the elimination of the low wavelet coefficients (variance of the residual noise Wiener = 202 versus 307).The resulting images has a good performance in terms of SNR for both method. Discussion/Conclusion: The fact that wavelet schemes tend to concentrate energy near edge features makes the result rather different than that found in standard Fourier based approach. We exploited the multiresolution nature of the wavelet representation to increase the accuracy of class boundaries. This approach appeared very promising for tissue segmentation especially in heterogeneous tumors. Mallat S. 1998. A wavelet tour of signal processing. Academic Press.

454 NLSnake: an active contour model for vascular wall quantification of multicontrast MR images C. Mansard Desbleds, M. Orkisz, B. Neyran, A. Anwander, L. Chaabane, E. P. Canet Soulas, I. E. Magnin, P. C. Douek; Creatis UMR 5515, CNRS, Villeurbanne, FRANCE. Introduction: Increasing number of studies are using high-resolution MRI to evaluate vessel wall abnormalities. However, image analysis is still either purely qualitative or manually performed without any dedicated tool. NLSnake, an active contour model has been optimized to segment and quantify MR images of the vessel wall. We present the validation and the evaluation of this new tool on high-resolution MR images of rabbit aorta with multicontrast. Methods: An active contour or snake is a curve evolving from an initial shape towards a final solution, under external-forces action and internal-forces reaction. Advantage of snakes is its great adaptability, but the result accuracy often depends on initialization and parameters adjustments. The originalities of NLSnake are : its independence from the initialization, simplified parameter tuning, fast convergence and minimum user interaction. High-resolution MR images of thoracic aorta of WHHL rabbits were acquired in vivo at 1.5T (Vision, Siemens) with a T1 SE sequence (TR/TE=855/20ms, voxel=0.2×0.2×2mm3) and a FSE sequence for T2 and proton density (TR/TE=2500/54ms and TR/TE=2500/15ms, pixel=390µm and 310µm, respectively). All MR images were acquired with fat saturation and ECG triggering. After sacrifice, in vitro HR MR imaging of the aorta segment was performed. T1 and T2 weighted images were acquired at 2T with a SE sequence(TR/TE=600/21ms and TR/TE=1800/50ms, respectively, pixel=78-97µm, slice thickness=0.8-1mm). Performances of NLSnake (reproducibility and accuracy) on contour detection were compared to manual measurements. Results: For in vitro MRI, a very good correlation was obtained between NLSnake and the contours traced by the experts (slope of the regression 1.05, r=0.88, n=121 slices), with a maximum distance between NLSnake and expert-traced contours of 0.27mm. Variability between NLSnake and the experts was less than the inter-observer variability, while intra-variability of NLSnake was 1.2%. In vivo, contours by NLSnake were obtained on images with multiple contrasts, even with a high noise level (Figure). Conclusion: NLSnake is an optimized active contour model with fast convergence independent from pixel initialization and was successfully applied to quantify the vessel wall in various types of vascular MR images acquired in vitro and in vivo in animals. An interesting aspect of the developed tool the minimum user-interaction, the timesaving and the measurements accuracy compared to manual contour tracing. This tool is currently evaluated on MR images of carotid arteries in patients.

Figure: Inner and outer contours of the aorta on T1(left) and T2(right) weighted in vivo images of a rabbit aorta.

Imaging: Processing and Quantification 455 Cardiac perfusion classification from contrast enhanced MRI: An automatic approach based on principal components analysis M. F. Santarelli1, V. Positano1, A. Pingitore1, M. Panzarella1, B. Favilli1, L. Landini2, M. Lombardi1; 1MRI Lab, CNR Institute of Clinical Physiology, Pisa, ITALY, 2Dept. of Information Engineering, University of Pisa, Pisa, ITALY. Introduction: Cardiac Magnetic Resonance imaging results to be an efficient investigation method for myocardial perfusion assessment. In particular, from the analysis of contrast enhanced myocardial images it is possible to accurately assess the extent of injury after myocardial infarction and the presence of myocardial viability. Commonly, perfusion analysis is performed in a substantially empiric way by visual and parametric analysis of signal intensity time curves, with all the limitations that this approach causes versus reproducibility and reliability. So that, the development of a fully automatic approach is desirable. Principal Components Analysis (PCA) method has potential for transforming a multidimensional problem, i.e. residual curves data, in a lower dimensional domain, that is particularly suitable for data clustering. So that, the method we propose consists on a series of automatic postprocessing operations, including PCA, that allows an automatic discrimination of normal from pathologic regions. Methods: Data relevant to 2 normal volunteers and 2 patients with CAD have been acquired and analysed. Images have been acquired using a General Electric MR machine, at 1.5T (Signa Horizon CV/i), with a 4-elements phased array surface coil. Gd-DTPA (Omniscan, 2ml/10Kg) has been used as tracer. ECG triggered images are acquired, in diastolic phase, and Fast Gradient Echo–Echo Train (TE=1.3ms, FA=25°, FOV=38x28.5cm, matrix=128x128, 3 slices for 60 temporal frames) sequence has been used. In order to test the proposed method validity, firstly the acquired images have been analysed by cardiology and radiology specialist in order to determine the hypoperfused myocardial regions. Then, acquired images undergone to postprocessing operations for image registration, filtering and segmentation, that allow to automatically detect myocardial regions of interest (ROI) covering all the wall and to extract perfusion residual curves [1, 2]. Totally, 64 residual curves have been derived and they undergone PCA in order to automatically discriminate normal and pathologic regions.

Results: In figure, PCA results are shown, relevant to all acquired data; white squares are relevant to data obtained from normal regions and black triangles from pathologic ones.

S243

Conclusion: As the figure shows, the method proposed allows to discriminate quite well normal and pathologic data regions; so that, from preliminary results, it seems to be appealing as a full automatic method for accurate analysis of the pathology entity and extension. Refernces: 1. M.F. Santarelli, V. Positano, A. Pingitore, L. Landini, M. Lombardi, A. Benassi. Proc. 10th Int. Soc. Magn. Res. Med., 2002, p.2462. 2. V. Positano, M.F. Santarelli, L. Landini. EURASIP J. Applied Signal Processing. Vol 5 (2003); 413-421.

456 The relationship between contrast agent concentration and signal intensity on T1 weighted images for measuring perfusion with MRI M. Nazarpoor1, A. R. Moody2, A. L. Martel1, P. S. Morgan1; 1Academic Radiology, University of Nottingham, Nottingham, UNITED KINGDOM, 2Dept Medical Imaging, Sunnybrook and Women's Health Sciences Centre, Toronto, ON, CANADA. Introduction: The relaxation time T1 depends on the concentration of paramagnetic contrast agent. To calculate perfusion parameters from dynamic contrast enhanced MRI acquisitions, the concentration must be measured; however measuring T1 with high temporal resolution is difficult. At low concentrations, the relationship between changes in 1/T1 and concentration can be considered to be linear. To maximise the concentration, and hence signal to noise in perfusion images, the range of this linearity must be known. This work studied the linearity using inversion recovery (IR) TurboFLASH sequences, and the effect of inversion time (TI). Methods: To assess the relationship between signal intensity (SI) and concentration, a water-filled phantom containing vials of different concentrations of Gd-DTPA (0-7.9mmol/L) was used. Mean SI was obtained in ROIs using T1-weighted images. All studies were carried out using a 1.5T clinical MR scanner with a standard head and neck coil. Coil non-uniformity was corrected on SI. A simple relationship giving the SI for IR sequence is [1]

Where S(t) and S0 are SI after and before, and T1pre is relaxation time before injection of contrast agent, TR is the image repetition time and K is a constant. This gives the relationship between concentration, C(t), and SI, and was fitted to measured data. To calculate the portion of the curve that could be considered linear, the R2=0.95 was calculated from correlation of the first n points for each TI. TI was varied between 400 and 800ms. Results: Figure 1 shows a typical result from SI versus concentration (TI=800ms). Figure 2 shows the variation of the maximum concentration of the linear range with TI. Discussion: Different papers [2,3] have reported that there is a linear relationship between SI and concentration up to a concentration of 0.8 or 1mmol/L for the IR sequence. Our findings indicate that the linearity not only is dependent on concentration but also it is dependent on TI. A typical result at TI=800ms shows the linearity up to 2.5mmol/L. The difference between the present study and previous reports may be due to using different TI and possibly the method of measuring the maximum linearity. We can use about 2 times more than the previous concentration of contrast agent for IR

S244

Imaging: Processing and Quantification

at TI=800ms for measuring the blood flow on T1-weighted image [2], resulting in improved perfusion images. References: 1. Hornak [2000],The basic of MRI, Chapter8 2. Moody et al.[2000],Investigative Radiology 35:01-411 3. Canet et al.[1995],JMRI 4:411-415

demonstrated in the liquid crystal sodium-linoleate in water[2]. Materials and Methods: Sodium-linoleate in water was prepared by mixing equimolar quantities of 99% linoleic-acid (SigmaAldrich, Germany) and 1.0N NaOH. NaCl solutions were used as references for the experiment with the crystal and volunteer experiments performed in the human calf. Images were acquired in a clinical 1.5T Symphony (Siemens, Germany) with a home-built 23Na saddle-coil. A FLASH sequence with TR=100ms, TE=3.3ms and 20mm slice-thickness was used. The RF amplitude was varied between 10-140V. Regions-of-interest were selected and the data were fit to eq(1). The resulting frequency w was used to fit a simple sinusoidal function to the saline data. The percentage of signal associated with the FAE is:

(2) The measured T1 were 35ms for the saline solution and 8ms for the crystal. Results and Conclusions: Figure 1(a)-(b) show the plots of the signal intensity as a function of RF amplitude and the fits to these data. The %FAE calculated from the fit is 68% for the muscle and 70% for the crystal, in agreement with the result of 77% from[2]. The fits describe the measured data accurately and the experiments were reproduced in 5 volunteers, which indicates the presence of 23Na quadrupolar-splitting in the human calf muscle. Therefore, it has been shown that 23Na quadrupolar-splitting can be measured with MRI.

457 23Na Quadrupolar Splitting in the human calf in-vivo using 23Na NMR imaging S. Nielles-Vallespin, M. Bock, A. Bankamp, R. Umathum, L. R. Schad; Biophysik und Medizinische Strahlenphysik, Deutsches Krebsforschungzentrum, Heidelberg, GERMANY. Introduction: The signal as a function of RF-pulse amplitude is analysed using MRI, to investigate the presence of 23Na quadrupolar-splitting in-vivo. 23Na is a spin-3/2 nucleus with an electric quadrupole moment[1]. If the sodium ions are bound to macroscopically ordered structures, quadrupolar splitting of the energy levels can occur. Joseph and Summers[2] proposed a method based on the prediction that the nutation rate of the unshifted inner spectral line is twice that of a system without quadrupolar splitting. A voxel containing sodium both with and without quadrupolar-splitting could be characterised by a partial-volume approach:

(1) where S1 and S2 are the signals from the unsplit and split components respectively, u is the RF amplitude and w is the frequency of the sinusoidal function. This flip-angle effect (FAE) has been

Figure 1. Signal versus RF-amplitude: (a) sodium-linoleate and saline, (b) calf muscle of a volunteer and saline. 1H(c) and 23NaFLASH(d) image of the calf and saline reference. References: 1. Duvvuri et al.(1999) JMRI.9:391-394 [2]Joseph and Summers(1987) MRM.4:67-77

Interventional, Safety, Bioeffects

Interventional, Safety, Bioeffects 458 Passive visualisation of catheter extruded with dysprosium - a promising new approach S. R. Hegde1, M. E. Miquel1, R. Birtchnell1, V. Muthurangu1, A. M. Taylor1, S. F. Keevil1, D. L. Hill1, R. Razavi1,2; 1Cardiac MR Research Group,Division of Imaging Sciences, King's College,5th Floor,Thomas Guy House,Guy's Hospital, London, UNITED KINGDOM, 2Department of Congenital Heart Disease, Guy's Hospital, London, UNITED KINGDOM. Introduction: MR guided cardiac catheterisation is now possible in patients. [1]. In our combined x-ray and MRI (XMR) suite 14 patients have undergone MR guided diagnostic cardiac catheterisation thus far. We used fast imaging protocols and passive contrast mechanisms to image Swan-Ganz catheters with the balloon at the tip filled with CO2.The tip of the catheter was therefore clearly seen as a signal void in a steady state free precession (SSFP) sequence.However, we were unable to visualise the rest of the catheter and hence felt the need to improve on passive visualisation techniques and design a catheter that is better visualised under MRI Methods: Catheter design: Catheter tubing was extruded with dysprosium oxide (prototype provided by NuMed, Inc, Hopkinton, New York, USA). We chose dysprosium oxide as it has been shown previously to be well suited for susceptibility-based visualisation [2]. We manipulated the catheter in a flow- phantom to simulate endovascular intervention. Phantom set-up: Fig 1 1. Pump 2. 'Heart' phantom (26.5 x26 x10) 3. Perspex box with Gelatin (12 x6 x6) 4. Reservoir 5. Outflow tube 6. Portal for catheter insertion. The phantom (Fig 1) consists of an acrylic box filled with gelatin (Aldrich Chemical Co Ltd, Gillingham, UK) in which is embedded a ‘heart' phantom with inflow tube from a pulsatile blood pump (model 1423, Harvard Apparatus, MA, USA) and an outflow tube connected to a 5 -litre reservoir of water doped with manganese chloride (20mg/L) to simulate the relaxation properties of blood (T1~1000ms,T2 ~120ms). Stroke volumes of 60ml and a heart rate of 70bpm were used. The scanning was carried out in a 1.5T Philips Intera I/T MRI scanner. Results: Fig 2 Fig 3 Fig 4 For imaging (Fig 4) we used real time SSFP sequence (B-TFE, TR: 2.6 ms, TE: 1.3ms matrix: 192x256,FOV: 400 mm, slice thickness: 15mm,temporal resolution ~ 9 f/s). We were able to successfully visualise and manipulate the catheter. Conclusion: We have shown by this technique that the catheter properties can be adapted to enable susceptibility-based visualisation of the whole length of a catheter regardless of orientation to the magnetic field. However, more work needs to be done to ascertain the biocompatibility of this catheter. References: 1. Razavi R et al. Clinical use of XMR in the diagnosis and treatment of Congenital Heart Disease. Journal of Cardiovascular Magnetic Resonance Vol 5, pp. 1-2.

S245

2. Bakker CJ, Hoogeveen RM, et al. MR-guided endovascular interventions: susceptibility-based catheter and near-real-time imaging technique. Radiology 1997; 202:273-276

S246

Interventional, Safety, Bioeffects Results and Discussion: The controller algorithm was tested exvivo (fresh meat samples) and in-vivo (rabbit thigh). Figure 2: example of temperature evolution at the centre of the trajectory in-vivo (14mm x 11mm elliptic ROI). Seven successive can be identified on the temperature time course. Dotted line indicates the target temperature (50°C). The feed back loop showed good stability against the experimental noise in MR thermometry. A fast convergence to the objective temperature was obtained, together with good spatial uniformity (less than 10% fluctuations). Conclusion: Coupling of focused ultrasound and MR thermometry, combined with a temperature regulation algorithm, can provide spatial-and-temporal control of the temperature within an elliptic ROI, which is much larger than the dimension of the elementary focal region. References: 1. Salomir et al,JMRI,12 :571-583;2000 2. Palussière et al,MRM,49:89-98(2003)

459 Focused ultrasound guided by Magnetic Resonance Imaging: Spatial-and-temporal control of temperature evolution during local hyperthermia R. Salomir1, C. Mougenot1, J. Palussiere2, N. Grenier3, C. Moonen1; 1Imagerie Moléculaire et Fonctionnelle, Université de Bordeaux 2, Bordeaux, FRANCE, 2Radiologie, Institut Bergonié, Bordeaux, FRANCE, 3Radiologie, CHU Bordeaux Pellegrin, Bordeaux, FRANCE. Introduction: Of the different modalities to induce local hyperthermia, focused ultrasound is the only non-invasive technology available at the moment. MRI can provide real-time thermometry and allows on-line, automatic control of temperature evolution. Treatment of a large tissue volume can be achieved by moving the focal point along an inside-out spiral trajectory (1,2). In this study, a spatial-and-temporal controller is designed for the control of the temperature evolution within the target region. The aim is to reach a predefined temperature profile after a few successive trajectories. Materials and Methods: The objective temperature profile is taken proportional to the theoretical thermal build-up as it should be obtained in a homogeneous tissue, taking into account the heat diffusion (D) and using the time course of the focal point position and ultrasound power corresponding to the first passage (Figure 1.a). Starting with the second passage, the controller algorithm modifies the time course of the focal point displacement, using online provided MR-thermometry data, in order to adapt the local deposition of thermal energy. Figure 1b illustrates the modulation principle, using a slope-type modulation function (thin line). Thick line is the resulting temperature profile (simulation for D=0.05mm2/s). Experiments were performed on a Philips 1.5T Intera clinical scanner equipped with an MR compatible 14-ring spherical ultrasound transducer (R=80mm, f=1.5MHz), integrated in the bed of the MR system. The probe can be hydraulically moved in the horizontal plane and may describe a predefined trajectory independently of the MR scanner acquisition. 3D temperature maps were acquired each 3.6 sec, with a spatial resolution of 1mm x 1mm x 4mm, covering a volume of 128mm x 128mm x 48mm.

Interventional, Safety, Bioeffects

S247

460 Predictability of acoustic gradient noise in MR imaging K. Elsner1, C. Oesterle1, M. Bellemann2, J. Hennig1; 1Dept. of Diagnostic Radiology, Section of Medical Physics, University Hospital, Freiburg, GERMANY, 2Dept. of Medical engineering, University of Applied Science, Jena, GERMANY. Purpose: Fast MR imaging sequences suffer from high acoustic noise levels due to rapid switching gradients. The frequency response function (FRF) quantitatively describes the characteristic noise properties of the MR system. The theoretical acoustic spectrum (p) of a given gradient shape (g) is given by the product p = FRF * FT(g). The purpose of this work is to verify the predictability of acoustic noise (i.e., spectral distribution and sound pressure level [SPL]) in comparison with the measured acoustic noise. Methods: All acoustic measurements were performed on a Siemens 1.5 T Sonata system. Acoustic recordings were conducted with a Bruel & Kjaer type 2238 Mediator sound level meter and the sound software Cool Edit 2 Pro with a sample rate of 44100 Hz. 64 repetitions with TR = 500 ms were acquired. Using Matlab, the recorded acoustic data were normalized to the measured SPL, averaged, and then Fourier-transformed to generate complex-valued acoustic spectra. To obtain the FRF, the acoustic response to a broadband gradient pulse (0-8 kHz) was acquired. The resulting acoustic spectrum, normalized to the amplitude of the FT(g), is the complex-valued FRF. To compare the measured acoustic response to the theoretical acoustic spectrum (p), several simple gradient shapes (g) were evaluated. Results: Fig.1 shows the results for a trapezoidal test gradient in zaxis with linear ramp shape, amplitude = 17.5 mT/m, duration = 2.5 ms, and ramp time = 0.5 ms. Shown are in a) the FRF in the corresponding gradient axis and the FT(g), in b) the resulting theoretical acoustic spectrum, and in c) the measured acoustic spectrum. The measured and predicted spectral distributions are in good correspondence apart from the low frequency range (< 500 Hz). Tab.1 gives an overview of the different test gradient shapes and the resulting measured and predicted SPL values. Discussion: Further evaluation of the FRF has shown that the low frequency range is superimposed by peaks resulting from residual background noise (air condition). Thus, the predictability of the spectrum in this range is not as good as in the higher frequency range. This also affects the prediction of the SPL. Especially for smooth gradient shapes with relatively long ramp times, the corresponding FT is rather narrow. This means that the predominantly excited acoustic frequencies are in this low frequency range and thus the impact of erroneous background contributions is rather high.

461 Optical transmission of MRI signals: A safe alternative for internal MRI probes O. Memis1, O. Aytur1, E. Atalar1,2; 1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, TURKEY, 2Departments of Radiology and Biomedical Engineering, Johns Hopkins University, Baltimore, MD. Introduction: In the conventional MRI instruments, after echo signals are received by an MRI coil, they are transmitted through an ultra-low noise transmission system consisting of electrical cables. This design proved to be effective over the years, but the possibility of inducing high electrical currents on these electrical cables are safety concerns which prevented placement of internal MRI coils inside the body. Here we propose an alternative method, "optical transmission" of the MRI signal. Subjects and Methods: The proposed system consists of a receiver MRI coil, two-stage low-noise preamplifier and a low-noise laser diode connected to a photodetector with fiber optic cable in between. Decoupling in the antenna is achieved passively by two back-to-back PIN diodes, also providing protection to the circuitry. After the antenna, the ultra-low noise first-stage preamplifier (gain:24dB; noise figure:0.5dB) is followed by high gain second stage (gain:32.5dB) and connected to the pigtailed laser diode as shown in the figure. The fiber is connected to the photodetector

S248

Interventional, Safety, Bioeffects

where MRI signals are converted to electrical signals. Current power consumption is small enough to be operated by a battery.

Results: The overall noise figure of the system was measured to be <1dB which suggests that total SNR reduction in the images due to optical MR signal transmission system is less than 15%. The total power consumption of the system is around 400mW, enabling to system to be operational for 1 hour with a 0.4W-hr battery. Our calculations indicate that with current circuit elements, overall power consumption can be reduced to <0.1W. Discussion and Conclusion: Our preliminary design suffers from several problems. We didn't make any special effort in reduction of the size of the design. The sizes of the components used suggest that the system can be miniaturized and placed in a tube with a very small diameter. Also, we are trying to reduce the power consumption further. Instead of passive decoupling, active decoupling is a feasible alternative which requires an additional fiber optic cable transmitting the control signal from the scanner to the coil but currently it is not implemented. In our proposed design, the optical cables are used in connecting MR coils to the scanner hardware and therefore provide safety to the patient, as there is not danger of induction of electrical currents on these cables. We believe that this design will enable placement of internal coils into body cavities without providing any hazard to the patient.

462 Active deep brain stimulation during MRI: a feasibility study J. C. Georgi, S. Heiland; Division of Neuroradiology, University of Heidelberg Medical Center, Heidelberg, GERMANY. Introduction: Deep brain stimulation (DBS) became an established method in neurosurgery. About 30% of the treated patients do not experience any improvement. MRI-examinations like fMRI, perfusion and diffusion MRI, could be powerful tools to differentiate responders/non-responders and to find reasons for the difference. The benefit of these methods has not been available for patients that have active DBS because of safety-relevant and methodical concerns. Induced line surges could cause heating of brain tissue[1] and stimulation of neurons. To analyze these effects, induced voltage and temperature near the electrodes and their lead was measured under various conditions. Methods: The studies were performed on a Siemens Symphony Quantum 1.5T scanner. Measurements of the line surges and the temperature were conducted with a digital oscilloscope and a fiberoptic thermometer. The stimulating electrodes were placed in a spherical NaCl-solution filled phantom. The temperature probes were placed along the electrode and its lead. The induced voltage was measured at the lead between electrodes and stimulator (Medtronic 3625) (Fig.1). The following sequences run: GE-EPI, diff.-weighted SE-EPI, 3D-FLASH, TSE, SE.

Results: The measurement of induction didn’t show any line surges by switching gradients. In contrast, the RF pulses are clearly observable. Depending on the sequence and setup, the amplitudes ranged between about 1V and far more than 2kV (Fig.2). At a simulated defect of the lead sparks could be observed between the cable’s disconnected ends. The temperature-rise at the electrodes and the lead also depended very much on the RF-power and the setup. It laid between 0.1°C and 60°C (Fig.3). Discussion: The studies show that MRI-examinations in principle can be performed on patients with active DBS. When the lead is placed along the symmetry axis of the scanner heating of electrodes and lead cannot be observed. In the range of some Volts one doesn’t have to assume physiological reactions since the carrier frequency of the RF-puls (63MHz) is too high to excite action potentials of neurons[2]. When the lead is placed near the RF coil, however, one can observe extensive spontaneous heating and high induced voltage. These effects also depend on the applied RFpower. They can lead to serious hazzards for patients. Defects of lead and electrode have to be excluded. Before patient-examinations are carried out on other scanners the presented results should be reproduced for the particular setup. References: 1. Liu CY et al. [2000] JMRI 12:75-78 2. Deetjen P, Speckmann EJ [1999] "Physiologie". UrbanFischer, München

Motion, Artefacts, Quality Control

S249

Motion, Artefacts, Quality Control 464 Cardiac MR: Performance assessment parameters S. Khan, J. De Wilde; Bioengineering, Imperial College London, Kensington, UNITED KINGDOM.

463 Polar optimisation for image registration J. Manjon-Herrera1, F. Place1, L. Martí-Bonmatí2, M. Robles viejo1; 1Area de informatica medica BET, Universidad Politecnica de Valencia, Valencia, SPAIN, 2Radiology Department, Clinica Quiron Valencia, Valencia, SPAIN. Introduction: Image registration is a powerful tool on quantitative medical image analysis. Mutual information theory and correlation ratio similarity measures has been widely used on this field. In last years the quantity of images related with a patient has been increased due to the better acquisition systems and shorter acquisition times. In same applications likes FMRI we can talk around 1000 image per patient. Registration procedures based on gradient calculation (like Powell) and pyramidal optimisation are slow for clinical use. It’s necessary to speed up the registration procedure to get the necessary results on a acceptable time. Subjects and Methods: In this paper we present an image registration method based on the polar representation of the contour of the images which has a constant cost time independently of the final transformation. Our method has the following steps: 1. For each image get binary mask using a theresholding based on ISODATA algorithm. 2. Obtain the polar representation (angle, distance) of the external contour of each image using the centre of mass as origin of the coordinate system. 3. Obtain transformations using polar information (translation, rotation and scale): Translation: Displacement that align the two centre of masses. Rotation: Circular displacement of the moving image polar course that makes the correlation coefficient of the two courses maximum. Scale: It´s calculated dividing the media of two polar courses. 4. Finally, we use Correlation Ration to perform the final local registration. Results: The time using gradient and pyramidal approach depends on the transformation to be applied. Our method has a constant cost time cost, around 1 second per image (256x256 pixels). Discussion/Conclusion: This method works very well for those kind of images that has the same external contour (MR/MR or MR/CT). But probably scale factor works not properly on MR/PET. We are now working to extent the method to 3D.

Introduction: Numerous phantoms have been developed for evaluating cardiac scanners. Of these, few actually focus on testing the performance of the scanner from an image quality point of view. The major challenge lies in the development of a dynamic test object that simulates the motion of the heart and is also capable of simultaneously testing a range of parameters affected by motion The purpose of this work was to constrain the design of a dynamic cardiac MR phantom to the parameters that reflected the effects of a moving structure on the image quality. Method: As imaging of the heart involves imaging of a constantly moving structure, it is highly probable that conventional performance assessment tests are not adequate for localizing problems that are encountered whilst imaging a dynamic structure. In addition, artefacts from flowing blood also present a challenge. The parameters incorporated into this dynamic MR phantom will enable a more accurate means of testing the systems performance and must be of high sensitivity and specificity. Incorporated into this phantom are three sets of resolution bars, each with a spatial frequency of 0.7mm. This spatial frequency represents the minimum resolution required for coronary vasculature imaging. Ghosting tests have also been included in the phantom. A small ghosting bottle is placed in lower left quadrant. Ghosting tests provide an excellent representation of the effects of motion. Flow and cardiac motion was simulated with the use of the UHDC flow system (Quest Image Inc). The pump was programmed and reproducible waveforms were developed. These waveforms represent the cardiac blood flow. A simple balloon was used to represent one of the cardiac chambers. Results: The figure below represents the positioning chosen for the placement of the evaluation parameters, spatial resolution and ghosting. The area shaded in grey corresponds to the area of motion. The objective behind this design format is to test conventional parameters in the presence of motion. Conclusion: The primary aim behind the design of a cardiac MR

S250

Motion, Artefacts, Quality Control

phantom is to enable optimal performance assessment of cardiac MR scanners. Spatial resolution and ghosting tests were integrated into the phantom as these tests permit quantification of motion effects.

465 Rigid motion detection and artifact correction in MRI Y. Kim1, Y. Ryu1, S. Chung2, C. Oh1; 1Biomedical Engineering, Korea University, Seoul, REPUBLIC OF KOREA, 2Reaearch Center, Medinus Co., Yongin-Si, REPUBLIC OF KOREA. Introduction: The purpose of this paper is to propose a new motion artifact correction method. In MRI, the Fourier-transform-domain data is usually acquired in the k-space, from which spatial-domain image is reconstructed by inverse Fourier transform. We found that, under the assumption that there is one directional object movement along the readout gradient direction during the scanning, image artifacts due to the motion can be corrected by using the motion information obtained from the raw data. Both simulation and experimental results from a conventional MRI system are presented to show the utility of the proposed method. Theory and Method: The 2-D k-space data F(.)can be written as , [1] where f(.) is the image. With both x and y directional motions, [1] becomes . [2] With only the movement in the read-direction, e.g., in the y direction, we have a simplified form of [2] as , [3] where h i (y) is the amount of motion for the i-th encoding. After the read direction inverse Fourier transform, we have

, [4] where . Equation [4] can be rewritten as . [5] For rigid translational motions in the read direction, we can characterize the motion as [6] For any given data for the i-th encoding in equation [5], . [7] From [7], the original signal without motion artifact can be obtained as [8] Results: Simulation has performed on translational motion using a set of images with 10 different motion phases. Figure 1 is describing the translational motion of the phantom and Fig. 2 depicts the correction result and their 1-D transformed data, where corrected image on (b) has almost no artifact compared with the uncorrected image on (a).

Fig. 1

(a) (b) Fig. 2 Discussion: From the simulation and experimental results, we found that the proposed motion artifact correction method is useful in removing the blurring and ghost motion artifacts on the images. This method should be useful for the correction of the artifact from one-directonal rigid motion for the clinical MRI system. References: 1. Y. G. Kim, C. H. Oh, S. K. Kim, S. T. Chung, Y. J. Yang, Y. H. Kim, Correction of k-Space Mapping Errors in Phase Encoding Direction, ISMRM 9, p. 758, 2001. 2. Zang-Hee Cho, Joie P. Jones, Manhir Singh, Foundation of Medical Imaging, John Wiley, 1993

466 Quality control for diffusion MRI: conclusions and results drawn from 2 years of practice I. Delakis1, L. Moore2, J. P. De WIlde1, M. O. Leach2; 1Department of Bioengineering, Imperial College, London, UNITED KINGDOM, 2CRC Clinical Magnetic Resonance Research Group, The Institute of Cancer Research and The Royal Marsden NHS Trust, Sutton, UNITED KINGDOM. Introduction: There is a need for a systematic quality control procedure that monitors the diffusion performance of an MRI system, ensures the reliability of long-term diffusion studies and allows the correlation of multi-centre measurements. Several system factors affect the ADC measurement. For a two-point ADC measurement (using the non-weighted (S0) and a diffusion-weighted signal (Sb)) the ADC can drift by d(ADC) due to: -discrepancies(db) in the b-value:

-inter-scan signal intensity fluctuations(dS0) unrelated to diffusion weighting:

Moreover, if n is the noise level of the system, the ADC precision (St.Dev.(ADC)) within a region of interest can be expressed as:

Motion, Artefacts, Quality Control Methods/Materials: Test-solution: Equations [A] and [B] show that a material with high ADC and low S0 is optimal for detecting inter-scan signal intensity fluctuations and b-value discrepancies. Equation [C] also shows that physiologically relevant signal levels provide a realistic indication of ADC measurement precision. Two aqueous solutions were developed: CuSO4 (1.8 g/l) and sucrose (38% w/w), their properties shown in Table 1. Table 1: Properties of test-solutions o

-5

2

-5

2

o

solution ADC (21 C) (10 mm sec) ADC/dT (10 mm /sec C) T2 (ms) SNR Sucrose 97

3

102

28

CuSO4

5

150

49

206

Imaging protocol: Two b-values (500 sec/mm2 and 1000 sec/mm2) were used in the diffusion sequence. Diffusion was applied independently along the orthogonal directions and three ADC maps were acquired (ADCSlice, ADCRL, ADCAP). Data Analysis: A parameter, Q, was defined, describing the ratio of the temporal standard deviation of ADC to the precision of the ADC measurement within a ROI:

A second parameter, R, provides an indication of the directional variation of the ADC measurement:

Results/Discussion: Table 2 shows the quality control program data. For b=500 sec/mm2 the Q for sucrose is approximately half the Q for CuSO4. This result is consistent with Eq.[A] and the relationship between the ADC values of the test-solutions (ADCsucrose ~ 0.5 ADCCuSO4), indicating that the ADC fluctuation is likely to be dominated by b-value discrepancies. For b=1000 sec/mm2, the Q for sucrose is greater than half the Q for CuSO4. According to Eq.[B], sucrose displays higher sensitivity than CuSO4 to inter-signal fluctuations due to its low baseline signal. Thus, the relationship between the Q values may indicate that interscan signal fluctuations play a greater role in diffusion with higher b-values. The directional dependence of ADC is governed by the d(ADC) of Eq.[A]. Indeed, R is highest for CuSO4 and b=500 sec/mm2 and lowest for sucrose and b=1000 sec/mm2. Conclusion: A quality control program for diffusion-MRI was developed, based on analytical study of the effect of MR system factors on the ADC precision and accuracy. Table 2: Quality Control program data solution

b-value (sec/mm2)

Q

mean R

CuSO4

500

0.86

1.56

CuSO4

1000

0.74

0.92

Sucrose

500

0.39

0.32

Sucrose

1000

0.59

0.28

Notes: The ADC values of the quality control program were corrected to 21oC before being used for the calculation of Q and R.

S251

467 1H-MRS: Application of an Integrated Quality Assurance Protocol on High Field 3T MR System P. Berardi1, C. Bergamini1, J. Alvarez-Linera2, S. Oliete2; 1Servizio Fisica Sanitaria, Policlinico S.Orsola Malpighi, Bologna, ITALY, 2Departamento de Neuroimagen y Resonancia Magnética, Hospital Rúber Internacional, Madrid, SPAIN. Introduction: As Medical Physics Department, our experience on Quality Assurance in MRI started in 1992, when a QA protocol was implemented to monitor two 0.5 T MR systems from the Acceptance Test, during monthly checks, up to decommissioning in 1999. In 2000 the QA protocol was extended with 1H-MR spectroscopy to a 1.5 T MR system and after 3 years we have acquired a good knowledge also on quality control spectroscopy. Now our goal is to verify the protocol compliance with an high field MR system 3.0 Tesla. Subjects and Methods: The main aim of the QA protocol is to optimize diagnostic accuracy by checking several physical parameters. In fact QA should be effective and efficient by means of a simple and fast method to check the performance of MR system. The whole protocol has been carried out on 1.5 T GE Signa Horizon EchoSpeed in “Policlinico S.Orsola Malpighi – Bologna IT” and on 3.0 T GE Signa in “Hospital Ruber Internacional Resonancia Magnetica – Madrid ES”. A 18 cm diameter sphere filled with a metabolite solution placed at head coil isocenter has been used as test object and PRESS and STEAM as acquisition sequences. QA protocol include the following measures: a. Signal linearity: increasing voxel size. b. Signal uniformity: moving a fixed size voxel along x,y,z direction. c. Signal contamination: moving the test object along z-axis. d. Voxel position: acquiring an image of the voxel and comparing it with test object image e. SNR and signal peaks of spectra. Results: The data collected are processed to check the signal linearity and uniformity, the true volume localization, and signal to noise ratio in spectroscopy. First data confirm a good linearity versus voxel size. An acceptable variation of signal has been observed within voxel localization (sd<5%). Localization of test object along z-axis produced a low variation in signal intensity (sd<3%). Discussion: From our experience this QA protocol has proved to be a powerful tool to monitoring physical parameters in time and to compare performances of different MRS systems. Furthermore has shown good flexibility upon different equipment and low time consuming, and so it meets the needs of effectiveness and efficiency. This is already a work in progress and we are going to obtain a more significant database to compare the performances between 1.5 T and 3.0 T to have a good characterization of systems differences.

S252

Methodology: Other

Methodology: Other 468 MRI pulse programming via the standard parallel port H. C. Seton, J. M. S. Hutchison, F. Resmer; Department of BioMedical Physics and BioEngineering, University of Aberdeen, Aberdeen, UNITED KINGDOM. Introduction: Recent improvements in processor speed have encouraged the development of PC-based MRI pulse sequence control using off-the-shelf scientific data acquisition hardware [1]. We discuss an alternative approach in which pulse sequences are generated using inexpensive DAC and ADC chips, under direct computer control, via the standard parallel port. The port comprises 8 data lines (bi-directional), 5 status lines (inputs) and 4 address lines (outputs). The address lines allow control of up to 16 devices. Methods: The pulse programmer is based around a 233MHz RiscOS computer [2]. The external hardware is interfaced to the computer's standard 25-pin parallel port via an 8 m cable. Four dual 8-bit multiplying DACs generate scaled RF and gradient waveforms (AD7528 Analog Devices, USA). Quadrature-demodulated NMR signals are acquired using two 12-bit ADCs (LTC1276, Linear Technology, USA). The lower 8 bits of the ADC outputs are joined to the main data bus, while the upper 4 bits are read in via the status bus. A further two AD7528 DACs generate gradient and B0 offset signals for field shimming and adjustment. Finally, a data latch (74HC573) is used to generate TTL pulses for RF gating and phase control. Pulse sequences are written for the RiscOS single-task environment in Acorn Basic 6.0 programming code. Images are reconstructed using standard FFT algorithms, again written in Basic 6.0. Results: During calibration the various waveforms were observed with an oscilloscope while software delays were adjusted to optimise pulse durations, rise times and ADC rates. The pulse programmer was used to acquire signals from a 0.01 T MRI system of in-house design and construction. The figure below shows a train of 32 spin echoes, collected from a yellow pepper (capsicum) of outside diameter 8 cm, using a CPMG sequence (TE=45 ms). The data indicate a T2 value of approximately 190 ms.

The following image of the same sample, together with a marker tube, was acquired using a standard spin-echo imaging sequence: 1282 pixels (cropped), TE=7 ms, TR=2000 ms, 7 mm slice, 64 averages.

Conclusions: We have shown that MRI experiments can be controlled successfully via the standard parallel port of a singletasking RiscOS computer. We have performed initial trials which indicate that we should expect similar performance from a DOS PC executing compiled C/C++ code. References: 1. Wright SM et al (2002), MAGMA 13, 177-185 2. Originally manufactured by Acorn Computers Ltd, but now supplied by Castle Technology Ltd, UK.

469 High-resolution MRI for visualization of intraocular lens implants S. Olt1, W. Haigis2, N. X. Nguyen3, S. Huber3, T. Meigen2, P. M. Jakob1, A. Haase1; 1Department of Physics, University of Wuerzburg, Wuerzburg, GERMANY, 2University Eye Clinic Wuerzburg, University of Wuerzburg, Wuerzburg, GERMANY, 3University Eye Clinic Erlangen-Nuernberg, University of Erlangen, Erlangen, GERMANY. Introduction: During cataract surgery the natural eye lens is replaced with an rigid artificial intraocular lens(IOL). Recently, several types of accommodative IOLs were designed and clinically evaluated, but so far their exact mechanisms in vivo are unclear. The ability to measure the response of the natural lens to a stimulus by MRI has already been reported[1]. Here we demonstrate the feasibility of visualizing the accommodative IOL in patients using MRI, which is the first step for dynamic MRI during accommodation. Subjects/Methods: Healthy volunteers and a patient with an focus-shift type IOL were investigated on a 1.5T whole body scanner (Siemens, Vision) using an 8cm surface coil. Subjects were asked to keep their eyes closed, so that the lens and the ciliary muscle were in a resting (unaccommodated) position. A 3D CISS sequence (12.3ms/5.1ms/70°) provided a high contrast between the lens and the vitreous body and the anterior chamber respectively (scan time :3min22sec, nom. res.:0.48x0.48x1.0mm3). A diagram of the IOL (type 1CU, HumanOptics AG, Erlangen, Germany) is provided in fig 1. This design has hinged haptics by which the accommodative ciliary muscle contraction is supposed to be translated into a forward movement of the lens optics thus allowing near focusing. Results: Fig 2 shows the normal eye. The vitreous body (VB) and anterior chamber (AC) both appear very bright in the CISS-images (95-98% water) along with the highly perfused ciliary muscle

Methodology: Other (CM). In contrast, the crystaline lens (L) is hypointense. In fig 3 the eye with the acrylic focus-shift IOL is shown. The IOL can be identified between the anterior chamber and the vitreous body with its haptics linked to the ciliary muscle. The distance from the IOL to the front of the AC as well as the distance from the IOL to the posterior end of the VB are measurable with an accuracy of 0.5mm. These distances could serve as a marker to determine the position of the IOL in both accommodated as in rest positions. Discussion: Since June 2000 over 9000 patients have received the focus-shift IOL shown in fig 1 with most promising clinical results. Our first results using MRI show that accommodation-related displacements of focus-shift IOLs, which are expected to be of the order of 0.5 to 0.8 mm, should be detectable. Therefore, MRI could be the method of choice for understanding of the function of the focus-shift IOL. References: 1. Strenk et al, IOVS 1999;401:162-1169

S253

470 Time distortion transformation of sech pulses J. Shen; Mib, NIMH, Bethesda, MD. Introduction: Various ingenious methods have been proposed to improve the hyperbolic secant pulses. In many cases, however, the improvements on peak or overall RF power requirement were made at the expense of the inversion profile or vice versa. Here we proposed a time-distortion transformation method to improve the intrinsic adiabaticity of the hyperbolic secant pulse. This new method lowers the required B1 threshold without broadening the transition bands. Results: We define the time-distorted hyperbolic secant pulse as F1 = 2πvAsech(βT(1-τ')); F2 = 2πAtanh(βT(1-τ')) where τ' = τ'(τ). The sweeping on-resonance adiabaticity factor is defined to be Q(ω=F2) = 2πATv2F12/|dF2/dτ|. For the time-distorted hyperbolic secant pulses described above Q(ω=F2) = µv2(dτ/dτ') where µ is the phase factor of the original hyperbolic secant pulse. Here we propose a time-distortion function τ' such that dτ/dτ' = µ1/µ + (µ2µ1)τ/µ where µ1, µ2 are the start (τ = 0) and end (τ = 1) phase factors of the distorted pulse, µ the phase factor of the equivalent undistorted hyperbolic secant pulse. We have µ = (µ2-µ1)/ln(µ2/µ1). To facilitate comparison with the undistorted hyperbolic secant pulses we use a simple method to calculate theB1 threshold (B1min): Q(ω=F2) = µvmin2 = 3.0. Then we obtain a dimensionless factor B1min*Tp =ln(2.0/1%)/π describing B1min for different hyperbolic secant pulses. The inversion profile can be characterized using the ratio (F) of the inversion bandwidth at Mz = -0.95 to that at Mz = 0.95. Table 1 lists the figures-of-merit for several hyperbolic secant pulses with different phase factors. Fig. 1 shows the resultant time-distortion function for µ1 = 5.0, µ2 = 10.0, T = 5 ms. The phase factor for the equivalent undistorted hyperbolic secant pulse µ = 7.2. The amplitude modulation functions for the distorted (-) and the undistorted (-) hyperbolic pulses are shown in Fig 2. The numerically simulated inversion profiles are given in Fig. 3 (top: distorted with µ1 = 10.0, µ2 = 5.0; middle: undistorted with µ = 7.2; bottom: distorted with µ1 = 5.0, µ2 = 10.0) which clearly shows that the distorted hyperbolic secant pulse (µ1 = 5.0, µ2 = 10.0) reaches full inversion at much reduced B1 than the equivalent undistorted one. Discussion: In addition to the linear scheme analyzed here, various other functions can be used to specify the time-dependence of the sweeping on-resonance adiabaticity. Alternatively, the timedependence can be arbitrarily parameterized and numerically optimized. Table 1. Figure of merit for sech pulses phase factor bw * Tp B1min * Tp 4 13.5 5.85 5 16.9 6.54 6 20.2 7.16 7 23.6 7.73 7.2 24.3 7.84 7.2(distorted) 24.3 6.68 8 27.0 8.27 9 30.4 8.77 10 33.7 9.24 15 50.6 11.3 20 67.5 13.1

F 0.56 0.63 0.69 0.73 0.73 0.73 0.76 0.78 0.80 0.86 0.89

S254

Methodology: Other training material, to be distributed on CD-ROMs and on the Internet, consists of a workbook with detailed tasks and a large image database (IDB). Course guide, curricula, timetables, training tasks and other materials are currently under development. The training programme, of 80 days duration, is to be introduced in hospitals as work-linked training, and the module leads to the attainment of specified competencies. A digital dictionary of terms covering the field of medical physics in five languages (English, French, German, Italian, Swedish) will also be developed. Training topics: The MRI training module consists of 20 submodules, divided into approximately 50 tasks, covering the MRI unit and its environment, pulse sequences (basic and advanced), gel phantom design and manufacturing, MRI signal and contrast, kspace properties, image artefacts, MR angiography and flow quantification, MR spectroscopy, comprehensive MRI and MRS quality control/quality assurance (QC/QA) protocols, image file transport issues and MR image formats, post-processing (fMRI, perfusion, diffusion, etc.), patient and staff safety, MR compatibility tests, commissioning, site-planning and installation. The practical use of MRI/MRS equipment is a prerequisite of the entire training process, and the trainee learns to apply imaging protocols, collect results, calculate parameters and draw appropriate conclusions from observations. Practical activities are supplemented by simulation tools and post-processing software. The IDB contains approximately 1000 images illustrating MRI-related equipment, test objects, phantoms, common clinical applications, artefacts, QC/QA and installation procedures. Discussion: Medical physics trainees often have limited time for hands-on training, as medical imaging equipment can not be purchased solely for training purposes, and such equipment is also intensively used in routine clinical work. The present training material offers off-site familiarisation with MRI equipment, by use of the IDB and the simulation software, and this facilitates optimal use of actual scanner time. The present material is primarily intended for the initial training of young medical physics graduates, but it might also be useful to established physicists, for continuous professional development, and to other hospital professionals.

472 Micro-MRI of the mouse anatomy G. J. Strijkers, A. Vilanova i Bartroli, W. Mulder, E. Heijman, B. M. ter Haar Romeny, K. Nicolay; Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, NETHERLANDS.

471 European medical imaging technology training (EMIT): Development of a comprehensive MR physics training programme R. Wirestam, A. Simmons, F. Milano, F. Ståhlberg, S. Tabakov, C. Roberts; EMIT Project Consortium, King's College London GKTSM, London, UNITED KINGDOM. Introduction: Schemes for MRI training are currently scarce, although the clinical use of MRI increases every year and professional training in medical physics becomes more and more important. Within the Leonardo Da Vinci EU project European Medical Imaging Technology Training (EMIT), a consortium of European universities and hospitals has been established to produce a comprehensive MRI training programme. Project description: This project aims at the production of an MRI training programme primarily for medical physicists. The MRI

Introduction: Rapid developments in molecular biology have created extraordinary opportunities to understand the connection between specific genotype, structure, and function. The phenotyping of transgenic mouse models has become especially important, because in this way genetic alterations can be linked to morphological changes. Traditional histological methods typically involve sacrificing, fixating, sectioning, and staining, after which the slices can be photographed, a time consuming and invasive procedure. Magnetic resonance imaging is rapidly becoming an important alternative to this method, allowing the interrogation of intact, opaque organisms in three dimensions at relatively high resolution. The goal of this work was to explore the possibility to use high-resolution MRI as a histological instrument, and to use mathematical tools to analyze, segment and visualize the mouse anatomy. Materials and Methods: Healthy mice were sacrificed and perfusion fixated using 4% buffered formalin, either directly into the

Methodology: Other heart or via the vena cave in order to keep the heart intact. GdDTPA was added for controlled contrast and to lower the T1 relaxation time of the tissues, allowing faster measurements. MRI measurements were performed with a horizontal 9.5 cm bore, 270 MHz MRI scanner with a Varian imaging console. A 3 cm diameter birdcage RF coil was used for acquisition. MR images were acquired with a standard 3D spin-echo sequence. Raw MR data were transferred to a personal computer for post processing with EasyScil/EasyVision, an advanced image analysis software package made by Philips Medical Systems. The final images were made using a combination of thresholding, manual segmentation, filtering, and smoothing. Results: The figure shows raw MR images and segmentations of the mouse heart. The images were recorded with a data matrix of 2563 and an isotropic resolution of 100 micrometer. Similar images were obtained for the brain and the kidneys. The results clearly show that MRI can provide a suitable tool for measuring morphological structures in the mouse with reasonably high resolution. Advanced image analysis tools will have to be developed when subtle morphological changes caused by genetic modifications have to be detected. We aim at developing targeted contrast agents to increase the specificity of the method. Conclusions: We have presented a method for measuring and segmenting high-resolution micro-MRI images of the mouse.

473 Contrast enhanced MR angiography versus TOF MR angiography in intracranial vessels using PAT optimized phased array 8-channel head coil Ö. Özsarlak, P. M. Parizel, J. W. Van Goethem, A. M. De Schepper; Department of Radiology, University Hospital Antwerp, Antwerp, BELGIUM. Purpose: to compare the eight-channel phased-array and standard circularly polarized (CP) head coils, to compare the three-dimensional (3D) time-of-flight (TOF) and contrast enhanced MR angiography techniques, to compare the conventional background suppression and “water excitation” techniques, and to define the effects of parallel acquisition technique (PAT) in visualization of the intracranial arteries. Materials and Methods: The study group is divided in two subgroups: (1); five volunteers studied with CP head coil, (2); ten volunteers studied with eight-channel phased-array head coil. The following MR angiography techniques were obtained on each volunteer: (1); TOF MR angiography with conventional background suppression, (2); TOF MR angiography with water excitation for background suppression, (3); low-dose (0.5ml) gadolinium-enhanced TOF MR angiography with water excitation, (4); contrast enhanced (10ml) MR angiography with water excitation, (5); contrast enhanced (20ml) MR angiography with water excitation. The second group was also divided in two subgroups, (1); five volunteers studied without parallel acquisition technique, and (2); five volunteers studied with parallel acquisition technique with an acceleration factor of two. The studied parameters were visualization of branches of the carotid and vertebro-basilar arteries, image contrast, and background suppression.

S255

Results: “Water excitation” is a valuable background suppression technique. For all studied parameters, eight-channel phased-array head coil was superior to CP head coil. For central portion of the intracranial circulation, unenhanced TOF MR angiography with water excitation was the best technique. For the proximal branches, contrast enhanced MR angiography technique was the best technique. The best technique for the evaluation of distal branches is low-dose contrast-enhanced TOF MR angiography with water excitation using eight-channel phased-array head coil. PAT technique with a factor of two decreases the acquistion time to 58% of that obtained without PAT. Conclusions: The eight-channel phased-array head coil is superior to CP head coil. Background stationary tissue suppression is significantly better with “water excitation” technique. For visualizing the central portion (Circle of Willis) and proximal branches of the intracranial arteries, unenhanced angiograms are superior to lowdose contrast enhanced MR angiograms. Low-dose contrast administration is improved the visualization and assessment of distal branches on both coil types. The use of eight-channel phasedarray head coil allows to perform 3D time-of-flight MR angiography with parallel acquistion technique, which will save 48 % of the initial imaging time, with a same image quality.

S256

Flow Quantification

Flow Quantification 474 Cerebral spinal flow (CSF) in patients with communicating and posttraumatic hydrocephalus: phase-contrast MRI data N. V. Aroutunov; Neuroradiology, N.N. Burdenko Neurosurgery Institute, RAMS, Moscow, RUSSIAN FEDERATION. Objective: To determine and compare the range of cerebrospinal fluid (CSF) flows for healthy volunteers, for the patients with communicating and open posttraumatic hydrocephalus and to demonstrate the efficiency of phase-contrast MRI with cardio-gating for controlling results of ventriculoperitoneal shunting (VPS) (CSF dynamics evaluation) in patients with open posttraumatic hydrocephalus and in the nearest postoperative period. Material and Methods: All MR-examinations were performed using high-field MRI 1.5 T. MR study included: T1-W1, T2-WI, phase-contrast MRI in sagittal orientation and across cerebral aqueduct (AC). The values of linear and volume velocities of CSF in cerebral aqueduct and 3-d ventriculostoma were evaluated. We examined 24 patients (16 females, 8 males, aged 12-39) with hydrocephalus, 21 healthy volunteers (12 females, 9 males, aged 11-43) and 11 patients with posttraumatic open hydrocephalus. The values of linear velocities of CSF pulsation in AC were evaluated by cardiac gated 2D phase contrast MR data. The square of AC was evaluated by high resolution T1-WI. Then stroke volume (SV) was calculated as area under absolute value of volume velocity per cardiac cycle. All examinations of posttraumatic patients were performed before, immediately after shunting and during following 2 weeks in every 3-rd day. Discussion: We visualized and estimated the CSF passage through the third ventricule, aqueduct cerebry, the forth ventricule and cisterna magna. The stroke volumes at the level of aqueduct cerebry were 69 ml (35-104), 301ml (118 - 474) and 560 ml before surgery, 170 ml after surgery (p<0.05), respectfully for healthy volunteers, patients with communicating and posttraumatic hydrocephalus. Mean linear velocity amplitude (LVA) was (5.7 ±1.8 ) cm/c, (13.01± 6.71) cm/c and 22,1 cm/sec before and 12,3 cm/sec after shunting. Clinically positive reaction was marked. In our experience CSF pulsation considerably exceeding the normal value (P<0,001), prognose the favourable outcomes after surgery. Conclusion: Phase-contrast MRI quantitative evaluation of CSF pulsative flow associated with the hydrocephalus should help determine the response to shunting in communicating hydrocephalus. Phase-contrast MRI at the level cerebral aqueduct showed the decreasing of CSF pulsation after shunting for open posttraumatic hydrocephalus. These alterations remained for 2 weeks after surgery in all patients.

475 Accuracy and reproducibility in flow measurements using SENSE P. Thunberg1, M. Karlsson2, L. Wigström3; 1Department of Biomedical Engineering, Örebro, SWEDEN, 2Department of Biomedical Engineering, Linköping, SWEDEN, 3Department of Medicine and Care, Linköping, SWEDEN. Purpose/Introduction: The purpose of this study was to investigate how the accuracy and reproducibility in quantitative flow measurements is influenced when using SENSE (1) at different reduction factors.

Subjects and Methods: Compared to an acquisition for a full FOV (i.e. R=1), the standard deviation of the measured flow at a certain reduction factor is given by

where N is the number of pixels within the ROI, x and y the pixel coordinates within the ROI and g the geometry factor. The expression for the standard deviation were validated by computer simulations for different reduction factors (R=1,2,3,4) and velocities (0, 0.25×venc, 0.5×venc and 0.75×venc). Blood flow measurements were performed in the ascending and descending aorta in a healthy volunteer using an ECG triggered breath hold acquisition. A fourelement phased-array coil was used. Results: The results from the computer simulations showed that the reproducibility for the flow measurements can be predicted by the derived expression. The only discrepancies occurred at R=4. In figure 1 normalized histograms for 2400 flow measurements at R=4 are shown for the four different velocity alternatives. As the mean velocity within the ROI approaches venc the measured flow tends to have a broader distribution and a lower mean than the expected (sold curve)

The blood flow measurements (Figure 2) revealed identical flow curves at all reduction factors except for R=4 in the descending aorta where peak flow were underestimated. Discussion/Conclusion: Flow measurements using SENSE is possible even at high reduction factors. In areas where the geometry factor is large and velocities are close to venc, aliasing in single pixels (2) causes lowered flow measurements compared to the expected, as shown in Fig. 1. In the in vivo measurement the mean geometry factor in aorta ascending was 0.6±0.1, versus 4.3±0.4 in the descending aorta. The mean velocity in the ascending aorta was 0.6±0.1 m/s, and 1.0±0.3 m/s in the descending aorta. The higher velocities and geometry factors resulted in aliasing in some voxels in the descending aorta but not in the ascending aorta for R=4. An a priori knowledge of the geometry factor for the actual image plane would imply knowledge of how a certain reduction factor will influence the flow measurement. At higher reduction factors the venc should be increased accordingly to avoid aliasing due to the increased standard deviation. 1. Pruessmann KP, et al. [1999] Magn Reson Med 42(5):952-962 2. Thunberg P, et al. [2003] ISMRM Toronto.

Angiography 476 Inflow effect correction for fast T1 GRE MR sequence M. K. Ivancevic, MSc1, I. Zimine, MSc1, X. Montet, MD1, J. Hyacinthe, MSc2, D. Foxall, PhD3, F. Lazeyras, PhD1, J. Vallée, MD, PhD1; 1Radiology, Geneva University Hospital, Geneva, SWITZERLAND, 2Inserm, Université Joseph Fourier, Grenoble, FRANCE, 3MR Clinical Science, Philips Medical Systems, Cleveland, OH. Purpose/Introduction: Blood inflow across an imaging plane results in a flow-related signal enhancement. This can lead to an overestimation of the arterial input function (AIF) defined in the blood pool. A new, flow-adapted signal calibration was developed in a dynamic phantom, validated in a phantom bolus, and applied in patients. Subjects and Methods: Phantom study: An MR-compatible setup with a continuous flow generating pump was used to develop the flow-adapted signal calibration. Signal was measured for various velocities and contrast concentrations, and samples were taken from the closed circuit for measuring the static signal of the corresponding concentrations. Flow was measured with a phase contrast sequence. The phantom was images on a Philips Eclipse 1.5T MR system using T1 gradient echo sequences. To validate the calibration, it was applied on a contrast bolus injected into the open circuit and collected at the outlet. Patient study: Renal MR perfusion imaging was performed on 13 patients as a bolus test prior to MR angiography. Two slices were superimposed on the same location. In the first group of patients (n=7) both slices were acquired in diastole, and in the second (n=6) one slice was acquired in systole, and the other in diastole. The velocity at the time of the k-space center acquisition was considered to affect the signal intensity. The flow corrected calibration was applied on flow-affected slices acquired in systole. Results: Phantom study: In case of the RF-FAST sequence and long preparation time, the inflow effect resulted in a severe signal saturation for high velocities. The effect was reduced by using the FAST sequence and shorter preparation time. Converting signal to concentration with the static calibration overestimated the peak of the bolus 3.2 times (p=0.65). Patient study: No difference was observed in the peak of the bolus in slices acquired in diastole (p=0.3). In systole the peak of the bolus was overestimated 2.46 ± 0.3 times (p = 0.0083) in comparison with diastole. After flow correction this ratio was reduced to 1.19 ± 0.16 (p=0.3). Discussion/Conclusion: This study shows the importance of the inflow effect in MR perfusion imaging, and presents an efficient method for correcting it. Our flow-corrected calibration method presents an important clinical potential for improving the robustness and accuracy of the arterial input function determination in MR tissue perfusion quantification using contrast media.

S257

Angiography 477 Prospective comparison of Multidetector CT Angiography (MDCTA), contrast enhanced MR angiography (CE-MRA) and digital subtraction angiography (DSA) in patients with carotid artery stenosis F. Pediconi, C. Catalano, F. Fraioli, F. Venditti, M. Danti, P. Nardis, R. Passariello; Radiological Sciences, "La Sapienza" University, Rome, ITALY. Purpose: To prospectively compare MDCTA and CE-MRA with DSA in patients with suspected internal carotid artery stenosis. Subjects and Methods: 43 patients with carotid artery stenosis > 50% on ultrasound examination prospectively underwent MDCTA, CE-MRA and DSA. MSCTA was performed acquiring a volume from the aortic arch to the intracranial circulation administering 80 mL of contrast agent at 4 mL/s after determination of the delay time by a test bolus. CE-MRA was performed on a 1.5 T scanner; 4 continuous measurements were performed starting together with contrast administration (15 mL of Gd-DTPA at 2 mL/s). We analyzed the degree of stenosis based on NASCET criteria, the morphology of the plaque and the presence of tandem lesions. Three independent readers interpreted the images and statistical analysis was performed. The interobserver variability was calculated. Results: There was significant correlation between MDCTA, CEMRA and DSA for degree of stenosis. Carotid artery stenosis > 70 % was detected with high sensitivity and specificity: 100% and 100% , respectively, for MSCTA; 95% and 100% , respectively , for CE-MRA. Regarding plaque morphology better results were obtained by MDCTA, with a sensitivity in detecting ulcers of 96% as compared to 89 % of CE-MRA. No significant difference was seen for tandem lesions. The interobserver agreement was perfect and nearly identical for both modalities. Conclusions: There was a significant correlation between MDCTA, CE-MRA and DSA in the evaluation of degree of stenosis. MDCTA appears a fast and very accurate technique in the assessment of plaque morphology.

478 Gadobenate Dimeglumine (MULTIHANCE®) in contrastenhanced magnetic resonance angiography R. La Ferla1, M. Daprà1, G. Pirovano2, M. Kirchin1; 1World Wide Medical Affairs, Bracco Imaging SpA, Milan, ITALY, 2World Wide Medical Affairs, Bracco Diagnostics, Princeton, NJ. Objective: Gadobenate dimeglumine (Gd-BOPTA, MultiHance®) has high relaxivity compared to other gadolinium agents due to a capacity for weak and transient interaction with serum albumin. The studies described were conducted to establish the optimum dose of Gd-BOPTA among four different doses for contrast-enhanced MRA in three different vascular regions. Materials and Methods: Fourteen European centers participated in three different Phase II dose-ranging parallel-group clinical trials to evaluate four doses (0.025, 0.05, 0.1 and 0.2 mmol/kg) of GdBOPTA for MRA. A total of 392 patients were dosed: 161 for carotid MRA, 94 for renal/abdominal MRA, and 137 for pelvic MRA. All imaging was performed at 1.5 T. Three-dimensional contrast-enhanced MRA (CE–MRA; performed within 26 s with the following parameters: TR<7 ms; TE<2.8 ms; α=40°) was compared with unenhanced time-of-flight MRA (UE–MRA).

S258

Cardiac

Images from each study were evaluated in double-blind fashion by on-site investigators and separately by two independent, off-site blinded assessors. A diagnostic quality score between 0 (=poor) and 2 (=adequate diagnostic information) was awarded to each of nine vascular segments in each study. The sum of these scores (between 0 and 18) was a measure of the overall diagnostic quality of the examination. Results: Good diagnostic quality and good correlation between off-site readers was observed at all doses in each study. Increasing diagnostic quality with dose up to a dose of 0.1 mmol/kg was observed for each territory and, for the renal/abdominal and pelvic studies, significant improvement over UE–MRA was noted. Although 0.05 mmol/kg dose was frequently diagnostic, the 0.1 mmol/kg dose provided a greater and more consistent diagnostic gain relative to unenhanced images, at least for the renal/abdominal and the pelvic studies. A dose of 0.2 mmol/kg provided little or no additional benefit. Excellent image quality was also obtained for post-contrast carotid MRA images. However, this study was challenged by the already good diagnostic quality of the unenhanced (3D–MOTSA) images. On the other hand, the shorter examination time of the CE-MRA (< 15 s) compared to the UE-MRA (> 10 min) favoured of the former technique. Conclusion: The results demonstrate that a dose of 0.1 mmol/kg of Gd-BOPTA is the optimally effective dose for CE-MRA in each vascular territory investigated. The higher T1 relaxivity of GdBOPTA in blood (9.7 mM–1s–1) compared to conventional gadolinium agents (between 4.3 and 5.0 mM–1s–1) suggests that improved diagnostic performance would be achieved with GdBOPTA at equivalent dose.

Cardiac 479 Accelerated CArdiac Selective Excitation (CASE) imaging in parallel using SPACE RIP W. E. Kyriakos1,2, D. Mitsouras1, F. J. Rybicki1, R. V. Mulkern1; 1Radiology, Brigham and Women's Hospital, Boston, MA, 2Radiology, Children's Hopsital, Boston, MA. Introduction: Selective excitation methods that restrict signal along phase-encoding directions are currently investigated to decrease image artifacts in cardiac imaging. One such approach, termed CArdiac Selective Excitation (CASE), uses a reversed spiral trajectory to traverse the 2D excitation space in order to selectively excite a column encompassing the heart. In this work we show the result of combining such an approach with SPACE-RIP [1] parallel imaging in order to achieve faster rates with no aliasing. Methods: Parallel imaging techniques utilize signals from several independent receiver coils. SPACE-RIP first computes the 1D Fourier Transform of the k-space signal received in parallel in all the coils. In 2D imaging, this operation results in phase modulated projections of the target image, weighted by the coil sensitivity profiles and constitutes a set of linear equations for each column in the image, whereby the unknowns are the pixel values for that column [1]. SPACE-RIP allows for the sub-sampling of k-space in an arbitrary fashion and for finer conditioning of the reconstruction than SMASH and SENSE . We tested SPACE-RIP in combination with a modified fast spin echo (FSE) sequence in which the 90 excitation pulse was replaced with a 24-loop spiral 4.1 ms CASE pulse to excite a 5.5 cm radial column of tissue orthogonal to the slice selective refocusing pulses. In Figure 1, the left hand image is a fully en-

coded 2D FSE with CASE, acquired using a 4-element array coil. The middle image uses half of the k-space lines acquired by subsampling every other phase encode. The higher image noise is consistent with SMASH and SENSE; moreover, for higher acceleration factors, aliasing artifacts are expected. The right hand SPACE-RIP (acceleration factor of 2) image demonstrates lower noise. Its k-space pattern characterizes the spatial energy distribution, and in conjunction with matrix conditioning by eigenvalue thresholding, artifacts are minimized at both low and high levels of acceleration. Conclusion: This work shows that SPACE-RIP can be used in conjunction with CASE imaging to effectively improve temporal resolution and minimize aliasing artifacts and noise amplification using matrix inverse thresholding. 1. Kyriakos WE, Panych LP, Kacher DF, et al. MRM 2000;44:301-308 2. Sodickson DK, Manning WJ MRM 1997;38:591-603 3. Pruessmann KP, Weiger M, Sheidegger MB et al. MRM 1999;42:952-962

480 MR imaging for the evaluation of left ventricular mass and volume of endurance athlete’s heart: a comparative study with 3D live ultrasound (3D US) I. Carbone1, M. Francone1, C. Catalano1, M. Danti1, K. Lanciotti1, S. De Castro2, S. Caselli2, A. Pelliccia3, R. Passariello1; 1Università La Sapienza, Department of Radiological Sciences, Roma, ITALY, 2Università La Sapienza, Department of Cardiovascular and Respiratory Sciences, Roma, ITALY, 3Institute of Sport Science, Italian National Olympic Commitee, Roma, ITALY. Purpose: Regular and extensive endurance training leads to a cardiac hypertrophy. The aim of this study was to evaluate the left ventricular mass and volumes of endurance athletes with MR and 3D US. Materials and Methods: Twenty-two healthy well trained rowers (20 males 2 females) underwent MR and 3D US within the same day. A control group of 22 healthy untrained volunteers (19 males 3 females) was enrolled in the study. Cardiac MR was performed with a 1.5 T magnet (Magnetom Vision, Siemens). Images were acquired with the subject in the supine position, by applying ECGgated breath-hold sequences. Nine to 14 short axis CINE images, with a slice thickness of 10 mm and without any interslice gap were acquired for the evaluation of left ventricular mass and volumes. Quantitative analysis including end-dyastolic and end-systolic volume (EDS, ESV) ejection fraction (EF), and LV mass (LVM), was performed off-line using a dedicated software (ARGUS, Siemens) by two blinded radiologist. Three-D US left ventricular mesurements were assessed on short axis, and quantitative analysis was blindly performed by 2 cardiologists. The interobserver agreement and the results of either MR and 3D US were compared. Results: LV masses and Volumes were significally increased in endurance athletes: LV volumes and masses were greater by 41 +18% and 31+- 14% respectively in comparison with the healthy volunteers.The average athlete’s cardiac mass and volumes assessed by MR and 3D US were respectively: LV mass 202,76 g (MR), 212,44 g (3D US); LV end diastolic volume 207,41 ml (MR), 230,89 ml (3D US); LV end sistolic volume 86 ml (MR), 90 ml (3D US).The interobserver agreement was almost perfect for either MR (K= 0.88) and 3D US (K= 0.90).

Cardiac Conclusion: LV hypertrophy induced by regular and extensive endurance training is perfectly showed by MR. Three-D US seems to be a good modality in the evaluation of LV mass and volumes.

S259

Conclusions: The size of the areas of late enhancement is well related to ventricular function, in particular with ESVI and EF, where ESVI is an index of myocardial functionality, and the transmural extension of necrosis seems to be decisive in predicting the postinfarctual remodelling.

482 Multi-TE SSFSE black-blood sequence in a single breathhold for T2 mapping of the myocardium: preliminary results D. Mandry1, J. Escanye2, F. Odille1, P. Le Roux3, P. Marie2, J. Felblinger1; 1Pediatric Radiology, Children Hospital, Vandoeuvre les Nancy, FRANCE, 2Nuclear Medicine, Nancy University Hospital, Vandoeuvre les Nancy, FRANCE, 3Medical Systems, General Electric, Buc, FRANCE.

481 Extension of late enhancement MRI in the evaluation of left ventricular function in the first six months after acute myocardial infarction G. Puppini1, G. Destro2, S. Montemezzi1, F. Furlan1, G. Gasparini2, G. Gortenuti1; 1Radiology Department, Civil Major Hospital, Verona, ITALY, 2Cardiology Department, Civil Major Hospital, Verona, ITALY. Purpose/Introduction: To valuate the efficacy of MRI to evidence the infarct size (IS) and to predict the left ventricular function in the first six months after acute myocardial infarction (AMI). Subjects and Methods: 40 patients (35 males, mean age 59,4 years) underwent, on day 11,6±7 and at six months (mean 167 days) after a first reperfused AMI (25 anterior, 15 inferior), to MRI evaluation (Siemens, Magneton Symphony, 1,5T) of the heart. Trufi short axis sequences were performed from the base to apex of left ventricle (6 slices) and then were calculated the end-systolic volume (ESVI), the end-diastolic volume (EDVI) and the ejection fraction (EF). Twenty minutes after 0.1 mmol/kg Gd-DTPA IV inversion recovery sequences (IR) were performed at 3 cardiac level, base, medioventricular and apex, divided in 6 segments (S). Number of segments with late hyperenhancement (HE), transmural (TR) or subendocardial (EN) were computed. Total area (TR+EN) of HE, the infarct size (IS), was measured and expressed as a percent of LV. In same 3 slices the systolic parietal thickening for any S was defined normal, akinetic (AKI) or hipokinetic (HYPO) by two independent observers. Results: The analysis of the database of both examinations series showed that IS was well related with ESVI, EF, HYPO, AKI and TR (p<0.001), a little less with EN (p<0.002) and EDVI (p<0.05). Correlation of TR vs EDVI and EF was very high (p<0.001) whereas the EN was neither related to ESVI and EDVI, nor to EF. Finally the segments AKI were well related to ESVI and EF (p<0.001) and with EDVI too (p<0.002). The HYPO segments were strongly joined to EF (p<0.001), ESVI (p<0.003) and EDVI (p<0.01).

Purpose: Marie et al [1] have shown that T2 determination allows the noninvasive detection of myocardial edema occurring in heart transplant rejection before the stage of allograft dysfunction. Their method is based upon a 2IR FSE black blood sequence; six echoes with different TEs were acquired, requiring six breath hold. We evaluated a new multi-TE SSFSE black-blood sequence which allows the acquisition of different TEs in a single breath hold. Subjects and Methods: Seven patients or volunteers were imaged on a 1.5T Signa system (General Electric) equipped with 33mT/m gradients. A black-blood FSE sequence (2IRFSE) used routinely on patients at our site and a new single shot multiple-TE black-blood FSE sequence (multi TE-SSFSE) were compared to a spin echo sequence (SE). The multi echo time SSFSE sequence is obtained by shifting the encoding steps along an augmented echo train. This way the recovery of longitudinal magnetization is kept unchanged. ECG-gated images were acquired during diastole in the short axis of the heart. The parameters for 2IRFSE are the following: 9 successive acquisitions in breath hold, 256*192 matrix, 10 mm slice, ETL=16, effective TE=4-72ms, TR defined by the heart rate. For multi TE-SSFSE, 6 images (+3 dummy) were acquired in a single breath hold. The matrix was reduced to 128X128, the slice kept to 10 mm and TE covered a range of 24-74ms. Before T2 mapping, an automatic registration of the images was done on a home made program developed with MatlabTM. Results: Figure 1 (T2 mapping in a healthy volunteer, after registration with 2IRFSE (left) and multi-TE SSFSE (right)) shows examples of T2 mapping with both techniques.

Breast

S260

484 Magnetic Resonance Imaging (MRI) in detecting early breast implant rupture: report on 259 cases S. Montemezzi1, G. Barbazeni1, A. Battistoni2, A. Marchi3, G. Rigotti2, G. Gortenuti1; 1Servizio Radiologia, Civil Major Hospital, Verona, ITALY, 2Dipartimento di Chirurgia Plastica, Civil Major Hospital, Verona, ITALY, 3Servizio Radiologia, Dipartimento di Chirurgia Plastica, Verona, ITALY.

Mean T2 values (ms) obtained with both techniques for each patient are summarized in table 1. The correlation between the two techniques is quite good and fitting is more accurate with the new sequence. Table 1: mean T2 values (ms) obtained with 2IR FSE and multi-TE SSFSE black blood sequences

Patient 2IR FSE multi-TE SSFSE

1 59 63

2 75 74

3 59 61

4 55 66

5 59 72

6 62 62

7 67 62

Conclusion: Due to this new sequence, complete mapping is feasible in a single breath hold, where routinely used sequence requires several apneas. However, parameter optimization is necessary to increase spatial resolution and further experiments to define a new threshold for acute rejection in heart transplant, or in other myocardial affections such as infarction or myocarditis. Reference: 1. P.Y. Marie et al., Transplant Proceedings, 30 1933-1935 (1998)

Breast 483 Performances and limits of breast MRI N. Kadi, J. Chopier, C. Marsault; Radiology, Hopital TENON, Paris, FRANCE. Learning Objectives: We propose to precise the actual indications of this technique in the management of the breast diseases. Sensitivity of Breast MRI is high but it's specificity don't exceed 65 to 85 % in the literature. We could be more performant if we respect the indications of this exam and if we optimize the technical parameters. We analyze the semiology and the dynamic behaviour of each lesion and we always correlate the MRI results to the clinical and the radiological abnormalities. By the time we precise the most frequent difficulties (Young woman, Elderly woman taking a substitute hormonal treatment) and the methods to minimize them. We conclude that the MRI could be a performant exam if we respect the indications, we optimize the technical parameters and we include it in a global breast diagnosis discussion.

Introduction: To evaluate MRI diagnostic accuracy in detecting early implant rupture as well as silicone gel bleeding. Matherials and Methods: From November 2000 to December 2002, 259 women underwent MRI implant evaluation. MRI was performed by using the Magnetom Symphony 1.5 tesla (Siemens). The imaging protocol included fast spin echo T2 weighted images and Turbo Inversion Recovery weighted for silicone images. 201/259 unoperated women were evaluated by clinical and/or MRI follow-up. 58/259 patients underwent surgery and MRI findings were correlated to the surgical reports. Results: MRI diagnosed implant rupture in 49 cases. Surgery assessed implant rupture in 47/49 patients. Surgery showed hematoma between the external shell of the implant and the capsule in one of the two patients who resulted MRI false positive for implant rupture . MRI did not assess implant rupture in 210/259 patients of whom 8 underwent surgery which resulted negative in 5 cases while in 3 detected early gel bleeding. The retrospective analysis of those cases, detected as negative by MRI, showed loose implant shell, wich was assessed as radial fold, distinct from the capsula by a small silicone amount. MRI sensibility, specificity, positive predictive value, and negative predictive value was respectively 94%, 99%, 95.9%, and 98.5%. Conclusion: In our experience, MRI has been effective in detecting early implant rupture. MRI diagnostic accuracy may be improved by assessing early silicone gel bleeding as well as differentiating between silicone gel bleeding and initial implant rupture, but the latter is not clinically important because in both cases surgical removal of the implants is required.

485 Contrast-enhanced MRI of the breast in the differentiation between radial scar and carcinoma F. Pediconi, R. Occhiato, F. Venditti, F. Fraioli, V. Votta, C. Catalano, R. Passariello; Radiological Sciences, "La Sapienza" University, Rome, ITALY. Purpose: The purpose of this study was to assess the value of contrast enhanced Magnetic Resonance Mammography (CE-MRM) in the evaluation of patients with mammographic suspected radial scar. Subjects and Methods: Fifteen women with radial opacities and black star findings at mammography, preoperatively underwent CE-MR mammography. Examinations were performed with a 1.5 T magnet with a bilateral superficial coil using a fat saturated T2weighted fast spin-echo sequence and 3D dynamic T1-weighted FLASH sequence acquired pre- and 1, 2, 3, 4, 5 min post-contrast. Dynamic contrast medium behaviour of the detected lesion were evaluated. Mammography and CE-MR mammography findings were subsequently compared to histopathology results. Results: No enhancement at MR mammography was seen in benign lesions, as confirmed at pathology. CE-MR mammography

Breast confirmed the presence of radial scar in 11/11 patients and of malignant lesions in 3/4 cases;1 malignant lesion was classified as borderline (enhancing lesion with no definitively malignant curve behaviour). CE-MR mammography provided a specificity of 100%, sensitivity of 92% and accuracy of 93%. Conclusions: CE-MR mammography is superior to other imaging modalities in the differentiation between radial scar and carcinoma. Lesions enhancing at CE-MRM, although morphologically suggestive of radial scar at mammography, have to be removed surgically.

486 Gadobenate dimeglumine (Gd-BOPTA) vs Gadopentate dimeglumine (Gd-DTPA) for contrast-enhanced magnetic resonance mammography (CE-MRM): improvement in lesion detection F. Pediconi, R. Occhiato, F. Venditti, M. Danti, A. Roselli, C. Catalano, R. Passariello; Radiological Sciences, "La Sapienza" University, Rome, ITALY. Purpose: The purpose of this study was to compare CE-MRM with 0.1 mmol/Kg Gd-BOPTA to CE-MRM with 0.1 mmol/Kg GdDTPA in breast lesion. Subjects and Methods: 20 consecutive patients with mammographic and sonographic suspected breast tumor underwent CEMRM using a 3D FLASH T1-w images acquired pre- and 1, 2, 3, 4, 5 min post-contrast . Gd-DTPA and Gd-BOPTA enhanced MRM were performed within 72 hours using a standard dose of 0.1 mmol/Kg at a flow rate of 2 ml/s. Signal intensity and dynamic contrast medium behaviour of the detected lesion were evaluated. Separate and combined assessment of unenhanced, enhanced and subtracted images was performed blindly by two readers. Sensitivity, specificity and diagnostic accuracy were calculated for lesion detection. Histopathologic examination after surgical excision was performed in all patients. Results: One patient was excluded from the study due to motion artefacts in subtracted images. In the detection of mammary carcinomas we obtained a specificity of 75 %, sensitivity of 100% and accuracy of 100% with Gd-BOPTA enhanced MRM; a specificity of 60%, sensitivity 88% and accuracy of 90% with Gd-DTPA enhanced MRM. Additionally, Gd-BOPTA enhanced MRM allowed to detect 3 multifocal (3/3), 1 multicentric (1/1), 2 bilateral tumors (3/3) and 3 radial scars (3/3). Gd-DTPA enhanced MRM detected 2 multifocal (2/3), 1 multicentric (1/1), 1 bilateral tumors (1/2) and 3 radial scars (3/3). Conclusions: Greater diagnostic accuracy for breast lesion detection and characterization is achievable with 0.1 mmol/Kg GdBOPTA as compared to Gd-DTPA.

487 Correlation of rim enhancement in MRI and PET of a large breast cancer in response to chemotherapy. A delivery related phenomenon with potential implications for therapy? S. I. K. Semple1, T. W. Redpath2, F. J. Gilbert1, T. S. Ahearn1, A. E. Welch2; 1Radiology, University of Aberdeen, Aberdeen, UNITED KINGDOM, 2Bio-Medical Physics, University of Aberdeen, Aberdeen, UNITED KINGDOM. Introduction: Neo-adjuvant chemotherapy is used to treat locally advanced breast cancer (LABC). Dynamic contrast enhanced MRI (DCE-MRI) and PET are used to monitor response to treatment1,2. Materials and Methods: 1 woman (37) with LABC received 6

S261

cycles of Epirubicin/Docetaxel chemo. DCE-MRI and PET were performed pre-chemo, 20 days after 1st and 3rd courses, and postchemo (pre-surgery). MR: 3D of both breasts, nine slice 2D gradient-echo of lesion at α=60, 100 and 350 to calculate T1, 350 scan repeated over forty time points with temporal resolution of 10s, 0.2mmol/kg Gd-DTPA administered frame 5. PET: 10min trans/atten scan 60min after 5mCi injection of 18F-FDG, 10min static emission scan. Results: PET/MRI images from 1st and last visits are presented infig. 1. Volume and MRI signal enhancement were noted for each tumour region, and PET SUV calculated2. Patient exhibited partial response to chemotherapy. Discussion: Patient presented with grade 3 invasive ductal carcinoma (IDC) with MRI/PET peripheral enhancement3 (fig. 1, table 1). Tumour center pre-contrast T1 equalled that of tumour periphery (~1,000ms). Post-chemo images show central enhancement /metabolic activity increasing (fig. 1, table 1). Tumour periphery SUV was initially high with lower central SUV. Through treatment, peripheral SUV reduced with central SUV increase. Although MRI and PET FDG uptake are based on entirely different mechanisms1,2, both display a similar initial lack of central uptake in these cases. The uniformity of T1 indicates the lack of central MR enhancement is not necrotic, whilst the lack of FDG uptake indicates that this is not due to central hypoxia4. Due to this correlation, we hypothesise that the initial lack of central enhancement and its subsequent increase is a delivery related phenomenon. One mechanism for this may be increased tumour pressure in large tumours5 compressing central capillaries, preventing passage of larger molecules. The phenomenon of increased enhancement post-chemo may be related to redistribution of blood flow to this existing, viable vasculature as a result of tumour pressure reduction. If contrast uptake in tumours is an analogue for drug delivery, the observed phenomenon shows that drugs may not be effectively delivered to the center of large tumours. References: 1. Knopp MV et al. MRI Clin North Am 1994; 2: 633-658. 2. Smith IC et al. J. Clin. Onc. 2000; 18(8): 1676-1688. 3. Matsubayashi R et al. Rad. 2000; 217: 841-848. 4. Guppy M. Biochem. Biophys. Res. Comm. 2002; 299: 676-680. 5. Ueki T et al. Cancer 2002; 95: 596-604. Table 1: MR analysis of tumour volumes (in cm3), % MR signal enhancement (post-pre*100/pre), SUV.

Visit1 Visit2 Visit3 Visit4

Tumour Volume 173.0 168.8 143.3 20.7

Central volume 38.3 7.8 0.8 0.3

Periph. enhanc. % 86 48 41 42

Central enhanc. % 1 6 28 106

Periph. SUV 0.149 0.045 0.036 0.035

Central SUV 0.069 0.029 0.078 0.097

S262

Lung 489 Influence of lung filling on hyperpolarised 3He transverse relaxation in rat lungs G. Guillot1, A. Vignaud1, L. De Rochefort1, E. Durand1, L. Darrasse1, D. Dupuich2, V. Stupar2, Y. Crémillieux2; 1U2r2m, CNRS UMR8081, Orsay, FRANCE, 2Laboratoire de RMN, CNRS UMR5012, Lyon, FRANCE.

Lung 488 MRI in staging of malignant pleural mesothelioma: correlation with surgical and pathological findings J. Maass, F. De Bei, M. Macagno, B. Vallone, M. Cassinis, G. Gandini; Diagnostic and Interventional Radiology, University of Turin, Torino, ITALY. Purpose: to evaluate the diagnostic efficacy of MRI in the presurgical staging of malignant pleural mesothelioma (MPM). Subjects and Methods: Beginning on February 99, fiftyseven consecutive patients with histological diagnosis of MPM, obtained by pleuroscopy, underwent MR examination. Nine patients were excluded from the study because of poor image quality. The results were calculated upon the remaining forty-eight patients. All examinations were reviewed by two readers and correlated to surgical and pathological findings. The four most important parameters in presurgical staging were independently analysed: invasion of the mediastinum, lymph-node metastasis, invasion of diaphragm or abdominal cavity, infiltration of thoracic wall. Three-plane T1-weighted breath-hold GRE sequences were used before and after intravenous Gd-DTPA; cardiac gating was limited to selected cases. Results: invasion of the mediastinum: sensitivity 75%, specificity 100%, diagnostic accuracy 82%; lymph-node metastasis: sensitivity 45,5%, specificity 100%, diagnostic accuracy 76,9%; invasion of diaphragm or abdominal cavity: sensitivity 94,7%, specificity 100%, diagnostic accuracy 96%; infiltration of thoracic wall: sensitivity 71,4%, specificity 94,7%, diagnostic accuracy 88,8%. Discussion/Conclusion: MR is a fundamental tool in presurgical evaluation of MPM. It is more sensitive than CT in cases of diaphragm involvement. CT is more reliable as regards lymph-node metastasis and invasion of bone structures or lung parenchyma. Optimisation of scanning technique and state of the art equipment are essential for good diagnostic features.

Introduction: Diffusion–weighted MRI using hyperpolarised gas has already proven sensitive to lung physiology[1]. In particular alveolar size has been correlated to ADC values in an animal model[2]. Transverse relaxation of hyperpolarised 3He at 1.5 T in human volunteers has been recently shown to be influenced by lung filling, more inflated lungs corresponding to longer T2CPMG values[3]. The purpose of this work was to obtain quantitative results of transverse relaxation in an animal model, with a better reproducibility of physiological conditions. Material and Methods: 10 Sprague Dawley rats, 6-week old were examined. They were anaesthetized, tracheotomized and disposed supine in the scanner. Pressure at the trachea was monitored during gas administration. Helium-3 was hyperpolarized by the spin-exchange method. MR measurements were performed on a 1.5 T scanner (Signa GE, with spectroscopic package). A coil matched to the animal size was used. The lungs were filled until a pressure of 30 cm of H2O (volume of about 8 ml) to reach inspiratory capacity (IC, A), then expiration was forced by extracting 10 ml of gas to reach a state of fully deflated lungs (B). A CPMG sequence was used without applied gradient, at an inter π-pulse spacing tCP of 20ms, with 128 echoes. The decay curve was recorded in conditions A and B for each animal, and the T2CPMG value obtained from a fit to a monoexponential. Results: The signal to noise on the decay curves was higher than 500. There was a systematic trend of longer T2CPMGs for more inflated lungs. The T2CPMG values (average and standard deviation) are shown in the Table. The reproducibility was better than 10%. The 2 conditions were differentiated with high significance (p<2.3×10–10). Discussion: These results confirm quantitatively the influence of lung filling on hyperpolarised–3He T2CPMG at 1.5 T, in the same way in rats as in humans. The IC T2CPMG value is shorter by 25% in rats than in humans. The alveolar size being about 2 times larger in humans than in rats [4], our observations confirm that T2CPMG increases with alveolar size. The relative sensitivity of ADC and of T2CPMG to lung filling should be compared. References: 1. Saam BT et al [2000] Magn Reson Med 44,174-179. 2. Crémillieux Y et al [2002] Proceedings 10th ISMRM Conference. 3. Vignaud A et al [2003] Proceedings 11th ISMRM Conference. 4. Mercer RR et al [1994] J Appl Physiol 77,1060-1066. Work supported by EC (PHIL, QLGI–2000–01559). Helium-3 T2CPMG in rat lungs for 2 lung filling conditions condition A [reproducibility] B [reproducibility] T2CPMG (ms) 145.4±10.6 [16] 98.5±11.0 [10]

Abdomen

Abdomen 490 Fat sparing of surrounding focal lesion cannot differentiate between benign and malignant lesion in patients with fatty liver L. Martí-Bonmatí, F. Peñaloza, E. Villarreal, M. J. Martinez; Radiology, Hospital Universitario Dr Peset, Valencia, SPAIN. Purpose: Perilesional fat sparing in liver tumors has been described mostly associated to metastasis (Chung J-J et al. AJR 2003;180:1347-1350). The association of perilesional fat sparing and hemangiomas is more rare (Terrier F. AJR 2000;174:869-870). We have evaluated the presence of perilesional fat sparing of focal hepatic lesion in patient with steatosis and its correlation with the benign and malignant origin of the tumor. Materials and Methods: From a large series of patients with liver tumors we have reviewed the MR images of 17 patients with fatty liver and focal hepatic lesions. There were 12 metastasis and 5 hemangiomas. The presence of a perilesional sparing ring, degree of liver steatosis, presence of biliary and portal vein obstruction, and lesion size were evaluated. Results: There were 10 females and 7 males (mean age 58±14 years-old). The degree of steatosis was graded as mild (10), moderate (6) or severe (1 case). A peritumoral sparing ring was found in 83% metastasis (5 partial and 5 complete) and 40% hemangiomas (1 partial and 1 complete). There was not a statistical difference in the presence of this ring between benign and malignant tumors (Chi-square p=0.2). Although lesions with a peritumoral ring were larger, the difference was not statistically significant (2.5 vs. 2.7 vs. 3.7, no ring vs. complete ring vs. partial ring respectively). There were 3 cases with a distal wedge shape sparing area, 2 cases with portal and 1 case with biliary obstruction. All these 6 cases were metastasic. Conclusions: Although usually associated with enlarging lesions, fat sparing surrounding a liver tumor can be seen in both benign (hemangiomas) and malignant (metastasis) lesions. The cause of peritumoral sparing is explained by the absence of portal blood flow to this area, most probably related to draining of arterial blood from the tumor and compression of hepatic venules in the surrounding parenchima by the tumor with decrease portal blood flow. Histopathologically, these areas have compressed hepatocytes cords, chronic inflammation and absence of fatty change.

491 Magnetic Resonance Imaging in staging of gastric cancer G. Stashuk, L. Portnoy; Moscow Regional Clinical and Research Institute, (MONIKI), Moscow, RUSSIAN FEDERATION. Objective: Studying the potency of magnetic resonance imaging (MRI) in pre-operative staging of gastric cancer. Materials and Methods: Results of MR Imaging of 63 patients (further operated on) were compared with histopathologic findings of resected gastric patterns. Apparatus “Signa-Contour” (General Electric, USA) with superconducting magnet and magnetic field induction of 0.5 T was applied for investigation. The sequences used are as follows: Single Shot Fast Spin Echo (SSFSE) – TE 70 ms, TR 814 ms, band width 15.63, FOV 45 cm, matrix 256x128, Nex 2, slice thickness 8.0 mm, spacing 2 mm; 7 sections were imaged during one breath hold (imaging time 16 s); Fast Spoiled GRASS (FSPGR) – TE 3.2 ms, TR 80 ms, flip angle 60°, band

S263

width 15.63, FOV 45 cm, matrix 256x128, Nex 1, slice thickness 8.0 mm, spacing 2 mm; 6 sections were imaged during one breath hold (time of imaging 13 s) and Fast Spoiled GRASS (FSPGR) out of phase – TE 6.8 ms, TR 80 ms, flip angle 60°, band width 15.63, FOV 45 cm, matrix 256x128, Nex 1, slice thickness 8.0 mm, spacing 2 mm; 6 sections were imaged during one breath hold (time of imaging 13 s). Cancer-infiltrated gastric wall image was obtained in 3 projections: axial, sagittal and coronal. The grade of serosal invasion of the stomach was classified according to the low intensity band around the cancer lesion. Results: MRI was non-specific in preoperative detection of gastric wall invasion and was based, first of all, on the degree of gastric wall thickening, its contour irregularity and extension of the lesion. Extraserosal tumor infiltration of gastric wall was visualized on spin-echo (SE) and, more reliably, on out of phase FSPGR images according to low intensity band seen between the wall the of waterfilled stomach and surrounding fat. Band disappearance or irregularity was considered as tumor invasion. MRI data concerning invasion degree corresponded with histopathologic findings in 16 of 19 (84%) pT2 patients, in 22 of 27 (81%) pT3 patients and in 14 of 17 (82%) pT4 patients. The mean accuracy formed 82%. Conclusions: MRI doesn’t allow to see all layers of gastric wall but is highly effective in tumor staging, and therefore may be used for pre-operative diagnosis and detecting the volume of surgical intervention.

492 Esophageal varices: MRI with an endo-esophageal probe L. Annet1, F. Peeters1, P. Starkel2, L. Hermoye1, Y. Horsmans2, B. E. Van Beers1; 1Radiology, St-Luc University Hospital, Brussels, BELGIUM, 2Gastroenterology, St-Luc University Hospital, Brussels, BELGIUM. Purpose: To assess the feasibility of using MRI with an endoesophageal probe to image esophageal varices. Materials and Methods: Twenty patients with esophageal varices proved by recent endoscopy were included in the study. MRI was performed with a receiver probe placed in the esophagus. Black blood, T2-weighted images of the varices were obtained with cardiac triggering and navigator gating of the right hemidiaphragm. The minimal diameter, maximal diameter, and area of the largest varix at MRI were recorded. Results: The esophageal varices were detected with MRI in each patient. Periesophageal varices were observed in 19 patients and paraesophageal varices in 14. The minimal diameter, maximal diameter, and area of the esophageal varices at MRI were significantly correlated with the endoscopic grade (r > 0.8, p < 0.001). Conclusion: MRI with an endo-esophageal probe is a feasible method to assess esophageal varices.

493 Single-session multi-organ MR imaging for the detection of metastatic disease spread T. Frattini1, P. Cipolla2, M. Kirchin2, A. Martegani1; 1Dept. of Radiology, Ospedale Valduce, Como, ITALY, 2WW Medical Affairs, Bracco Imaging SpA, Milan, ITALY. Purpose: The possibility to evaluate in a single imaging session the degree of metastatic disease spread from aggressive primary tumors may impact greatly on imaging time, costs and patient management. The study was performed to determine the feasibility for

S264

Abdomen

multi-organ (CNS, liver, pancreas and adrenals) lesion detection and characterization using an MR contrast agent with high in vivo relaxivity (r1=9.7 mM-1s-1) and dual renal and hepatobiliary elimination. Materials and Methods: 12 patients with known or suspected lung or pancreatic cancer underwent unenhanced T1- and T2weighted MRI of the liver and upper abdomen prior to receiving a single intravenous bolus injection of 0.15 mmol/kg gadobenate dimeglumine (Gd-BOPTA; MultiHance, Bracco Imaging SpA, Italy). Thereafter, dynamic T1-weighted images of the liver and upper abdomen were acquired followed by T1-weighted images of the CNS at ~15 min post-injection. Finally, delayed T1-weighted images of the liver were acquired at ~60 min post-injection. All images were evaluated for the presence of metastases. Assessments of diagnostic image quality and feasibility for single session multiorgan cancer staging with Gd-BOPTA was assessed using 4-point scales from 0=insufficient to 3=good. Results: Metastases were detected in more than one organ in 5/12 (42%) patients (liver and CNS in 2 patients; liver and adrenal glands in 2 patients; liver, CNS and adrenal glands in 1 patient). Four of the remaining 7 patients had lesions detected in one organ only (liver only in 2 patients, CNS only in 2 patients) while in 3 patients no lesions were detected. Among patients with detected liver lesions (n=7) additional lesions on delayed post-contrast images were detected in 2/7 (29%) patients. Maximum scores of 4 were given for 10/12 patients for both diagnostic image quality and protocol feasibility. Scores of 3 (satisfactory) were given for the remaining 2/12 patients. Conclusion: Single session multi-organ MR imaging for the detection of metastases is feasible with Gd-BOPTA. This may impact positively on both costs and MRI scheduling for those patients who would otherwise require more than one examination.

494 MRI and bile duct cysts: diagnosis and complications B. J. Op de Beeck1,2, M. van Weerelt2, K. Vanderdood2, V. Lambrecht1, M. Osteaux2, P. Parizel1; 1Radiology, University Hospital Antwerp UZA, Antwerp, BELGIUM, 2Radiology, AZ Free University Brussels, Brussels, BELGIUM. Purpose: Iconography of MR-findings in bile duct cysts and their complications. Marerials and Methods: A total of 8 patients with bile duct cysts underwent MRI. The examinations were performed between 1996 and 2002. All cases are surgically proven. MRI was done on a 1.5T unit (Magnetom Vision, Siemens, Erlangen; Germany) with phased array coil. The study included axial GRE T1-weighted and HASTE T2-weighted sequences. Additional breath-hold thick-slab RARE (30mm slice thickness) and multislice HASTE sequences (4 and 6mm slice thickness) were performed in the coronal and the paracoronal plane. Discussion: Although bile duct cysts are still rare conditions, their possible complications necessitate detection in a early state. A meticulous description of the anomaly is mandatory when surgical intervention is considered. The bile duct cysts are subdivided in five types according to the Todani modification of the Alonso-Lei classification. Complications of bile duct cysts include stone formation, pancreatitis, biliary cirrhosis, cyst rupture, liver abscess and cholangiocarcinoma. The relevant diagnostic clues on MRI are discussed and compared with our own experience. Conclusion: Although bile duct cysts are rare, it is important to detect them in an early stage. MRI offers the opportunity to visualize

the biliary system and is the screening method of choice. The knowledge of this anomaly is mandatory to prevent inadvertent injury during surgical or interventional procedures. Learning objectives: To recognize the range of imaging characteristics of bile duct cysts and their complications.

495 Usefulness of Secretin enhanced Magnetic Resonance Cholangio-Pancreatography for planning therapeutic endoscopies and follow-up studies after therapy E. Scarano1, R. Cantisani1, T. Cavallo1, M. Mutignani2, B. Lamorgese1; 1Radiology, S.Carlo Hospital, Potenza, ITALY, 2Surgery, A.Gemelli University Hospital, Rome, ITALY. Introduction: There is a moderate-high agreement of the secretinenhanced MRCP with ERCP to non-invasively assess the pancreatic duct. The secretin stimulation increases the caliber of the main pancreatic duct, especially in the tail and side branches improving their abnormalities and the detection of anatomic variants; furthermore this technique affords the functional evaluation of the pancreas. Subjects and Methods: MR studies were performed with a 1,5 T clinical imager (Horizon LX Echo Speed; GE Medical Systems) with use of a phased-array surface coil . Single-shot turbo spinecho T2 weighted dynamic MRCP was performed after orally administration of a negative contrast agent. S-MRCP was obtained before and 30-second intervals over 10 minutes after secretin stimulation in patients of group 1: with persistent hyperamylasemia (n=10), severe chronic pancreatitis ( n=12) idiopathic acute pancreatitis ( n=8) and in control group 2 ( n= 5). Morphologic features and diameter of the pancreatic duct were monitored before and during secretin stimulation. Duodenal filling volume was graded. Results: ductal diameters were significantly larger in 11 of 12 patients with severe chronic pancreatitis and pancreatic exocrine function parameters were significantly lower in this group. Ductal diameters were altered only in 2 of the other patients of group 1. Pancreas divisum was detected in 5 of 30 patients of group 1 (17%); Santorinicele was detected in one of these 5 patients, confirmed at ERCP. Therapeutic endoscopy was performed in seven of 30 patients ( 23%) of group 1 on indication of S- MRCP: 2 pancreatic duct stones after ESWL, 4 accessory papilla sphincteroplasty in pancreas divisum and 1 main duct sphincteroplasty for stenosis. Control S- MRCP was performed in all 7 patients. Six of seven patients were free of symptoms and showed an improvement of pancreatic exocrine reserve at control S-MRCP. Only in 1 patient stones cannot be removed endoscopically. Conclusions: in our experience the combination of MRI/MRCP with secretin stimulation prevents a number of diagnostic ERCP and influence therapeutic decision, moreover this technique allows specific estimation of pancreatic exocrine function.

496 The use of magnetic resonance cholangiopancreatography in iatrogenic bile duct injuries M. A. Aliev1, Y. A. Akhmetov2, S. E. Serikova2; 1Abdominal Surgery, Scientific center of surgery, Almaty, KAZAKHSTAN, 2Radiology, Scientific center of surgery, Almaty, KAZAKHSTAN. Purpose: to evaluate the role of MRCP in the assessing of iatrogenic bile duct injuries. Subjects and Methods: MRCP was performed with 1,5 T (Visart,

Abdomen Toshiba) superconducting magnet, programmed with 2D FASE and 3D FASE sequences in 12 postoperative patients suspected of bile duct injury duting the surgery. Results were compared with endoscopic retrograde cholangiopancreatography in eight patients, percutaneous transhepatic cholangiography in two, surgery in two. Results: MRCP was successful in all patients. Two patients had surgically proven bile duct excision injury on MRCP, and two other patients had stricture proven during percutaneous transhepatic cholangiography. One patients had findings suspected to cystic duct remnant`s leak on MRCP that was confirmed on ERCP. Conclusion: We can conclude, that MRCP accurately demonstrate postoperative biliary excision injuries and strictures and extremely useful for reparative surgery planning. Moreover, it also suggests the presence of cystic duct leaks in patients after the cholecystectomy. The method helps to avoid unnecessary ERCP in patients after laparoscopic cholecystectomy because of high diagnostic value in depiction of postoperative biliary tree.

S265

compared to controls. Conclusion: Lower molar concentrations of PDE, Pi and ATP in cirrhotic patients were found with dependence on Child-Pugh score which reflects degree of liver injury. Significant differences in MR spectra were found in various commonly diagnosed etiological groups (alcohol, viral and cholestatic). MR spectroscopy can be helpful in differential diagnosis in liver patients when etiology of liver disease remains unclear. Reference: Tosner Z, et al. [2001] Magma 13:40-46. Supported by grants IGA MZ CR 6630-3 and CEZ:L17/98:00023001. Absolute concentrations of 31P metabolites [mM] in patients with liver cirrhosis N

PME

PDE

ATP

All patients

49

3.06±1.18 1.22±0.58 (*)

6.32±2.64 ($)

2.70±0.94 (*)

CPS: A+B

26

3.15±1.06 1.23±0.59 (*)

6.95±2.42 (*,+) 2.78±0.95 (*)

CPS: C (severe)

23

2.95±1.32 1.20±0.57 (*)

5.67±2.75 (#,+) 2.61±0.94 ($)

497

Controls

9

2.78±1.27 1.69±0.70

9.88±2.69

Liver cirrhosis etiology by 31P MR spectroscopy M. Dezortova1, P. Taimr2, J. Sperl2, M. Hajek1; 1MR-Unit, Dept. Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague, CZECH REPUBLIC, 2Dept. Hepatogastroenterology, Institute for Clinical and Experimental Medicine, Prague, CZECH REPUBLIC.

CPS: C

Introduction: Liver cirrhosis is end-stage liver disease of various etiologies. Clinicians often need to find out the underlying disease and a degree of liver function derangement usually described by Child-Pugh score (CPS) to start an appropriate treatment. CPS is calculated from clinical and laboratory tests. An effort is made to obtain this information non-invasively, i.e. without liver biopsy. 31P MR spectroscopy enables the observation of liver metabolism in vivo through the signals of phosphomonoesters (PME), phosphodiesters (PDE), inorganic phosphate (Pi) and ATP. This study deals with absolute quantification of liver metabolites obtained in 31P MR spectra of patients with liver cirrhosis and the relationship between metabolite concentrations, clinical status and several etiological groups. Subjects and Methods: 49 patients (51.4±10.9 years) with liver cirrhosis of different etiology (alcoholic cirrhosis in 19 cases, viral hepatitis B or C in 16 cases, cholestatic liver disease - primary biliary cirrhosis, primary sclerosing cholangitis and biliary atresia - in 10 cases, and other etiology including 1 autoimmune hepatitis and 3 cryptogenic cases) and severity (CPS) and 9 healthy controls (41.8±11.0 years) were examined after overnight fasting. MR examination was performed on a Siemens Vision 1.5T using dual 31P/1H surface coil. Two-dimensional chemical shift imaging technique was applied in transversal plane: TR/TE=323/2.3 ms, FOV=480 mm, matrix 16x16, voxel volumes were 3x3x4 cm3, 12 acquisitions. Spectra were evaluated manually on a console and signal intensities of PME, PDE, Pi and βATP were used for the calculation of absolute concentrations. The methodology was published in [1]. Standard t-tests were used for statistical evaluation. Results: Table summarizes absolute concentrations of selected metabolites together with CPS. According to 31P MR spectra we can differentiate various etiologies of cirrhosis. Alcohol liver disease differed in all studied metabolites (p<0.05 for Pi and ATP, p<0.01 for PDE) from healthy volunteers. Patients with viral disease differed in PDE (p<0.001) and ATP (p<0.01) whereas patients with cholestatic disease had significant changes only in Pi (p<0.05)

Pi

(mild and moderate) 3.60±0.94

(*) p<0.05, ($) p<0.01, (#) p<0.001 from controls; (+) p<0.05 between CPS: A+B to

498 1H MRS assessment of hepatic steatosis in patients with chronic hepatitis C M. Krssak1, H. Hofer2, M. Meyerspeer3, E. Moser3, A. Lohninger4, P. Ferenci2, M. Roden1; 1Internal Medicine 3, University of Vienna, Wien, AUSTRIA, 2Internal Medicine 4, University of Vienna, Wien, AUSTRIA, 3Medical Physics, University of Vienna, Wien, AUSTRIA, 4Medical Chemistry, University of Vienna, Wien, AUSTRIA. Purpose: Histological grading of biopsies is the routine method for assessment of hepatic steatosis. Non-invasive 1H and 13C magnetic resonance spectroscopy (MRS) have already been applied for the quantitation of hepatocellular lipids (HCL) content in humans. Recent study in patients with hepatitis C suggested a negative correlation between HCL accumulation and the fibrosis grade, but the used method did not allow direct quantitation of HCL. The present study aims at validation of HCL measurement by 1H MRS with histological grading and biochemical analysis in patients with chronic hepatitis C. Individual water and lipid relaxation times and fibrotic changes of liver tissue will also be evaluated. Methods: HCL content was measured by localized 1H MRS on a 3T MR system in 31 patients with chronic hepatitis C at 2-4 hours before diagnostic liver biopsy. An echo time dependent series of breath-hold-triggered STEAM sequences (n=5; TE=10-70 ms; TM=30ms; NS=1) was applied using 10 cm surface coil. A VOI of 27 cm3 was placed in the lateral hepatic lobe. HCL content was quantified as ratio between lipid resonance intensity and the total signal intensity. T2 relaxation correction was performed with the individually measured water and lipid relaxation times. Histological grading of hepatic fat and fibrosis was performed by standard methods. Fat content is given as percentage of fatty cells in the biopsy. Morphological changes were graded according to five fibrosis grades. In a subgroup of 12 patients, total hepatic triglyceride (HTG) content was also measured biochemically by gas-liquid chromatography. Results: HCL content as assessed by MRS ranged from 1 to 38 % of the signal and from 0 to 80 % of fatty cells as assessed histologically. Both measures of steatosis were linearly correlated (r=0.70, p<0.001). Grade of fibrosis did not relate to HCL measured by

S266

Pelvis, GU

either MRS or histological grading. T2 relaxation times ranged from 20 to 81 ms for the water signal and from the 38 to 118 ms for the lipid signal. T2 relaxation times for water or lipid signal were not associated with hepatic fat accumulation and/or fibrosis grade. HTG content ranged from 1.7 to 60.3 mg/g dry weight and was linearly correlated with MRS-measurement (r=.63, p<0.03) but not with histology. Conclusion: In vivo 1H MRS fat quantification using an individual relaxation correction correlates with biochemical and histological assessment of liver fat content and can therefore be applied to quantify hepatic steatosis also in patients with chronic hepatitis C.

499 Liver fat content determined by proton magnetic resonance spectroscopy is increased in patients with type 2 diabetes R. J. H. Borra1,2, R. Parkkola1, R. Lautamäki2, P. Nuutila2,3, M. Komu1; 1Department of Radiology, Turku University Central Hospital, Turku, FINLAND, 2Turku PET Centre, University of Turku, Turku, FINLAND, 3Department of Internal Medicine, Turku University Central Hospital, Turku, FINLAND. Introduction: In vivo proton magnetic resonance spectroscopy (1H MRS) provides an elegant method for non-invasive assessment of the fat content of several tissues [1]. The purpose of our study is to develop a clinically feasible protocol for assessment of liver fat content and compare patients with type 2 diabetes to healthy subjects. Methods: A 1.5 T MR imager (Signa Horizon LX, GE Medical Systems, USA) with the flexible surface and body receiving coil was used for MRI and MRS of ten normal subjects (age 35±11 y.) and forty-three patients with type 2 diabetes (age 64±7 y., BMI 30±4 kg/m2, HbA1c 7,2±0,9%). The transverse T1W dual-echo FSPGR (out-in-phase) images were used for the definition of the centre and the dimensions of the single voxel PRESS MRS sequence (TR/TE = 3000/25 ms). The mean volume of the voxel was 29.9±6.4 cm3. Voxels were placed in the liver parenchyma outside of the area of the great vessels. The time domain fitting of the signal was performed using the jMRUI data analysis package [2]. The fat (methylene + methyl peak amplitudes) and water amplitudes were corrected due to different T2 decay [3] and molar concentrations of H1 nuclei in fat and water [1]. Liver fat content was defined as fat in relation to the total weight of liver tissue [3]. Results: The mean liver fat content was extremely significantly increased in diabetic patients (11.2 %) as compared to normal subjects (1.65 %) (t-test, p< 0.0001). A typical region for the voxel in diabetic patient liver MRS is displayed in the T1W image below with the corresponding spectrum.

Conclusions: 1H MRS provides an accurate tool for in vivo assessment of liver and fat content with many possible applications to diabetes research. The demonstrated possibility to obtain good quality

spectra with the body coil facilitates future clinical application of this technique. Aknowledgement: This research has been supported by a Marie Curie Fellowship of the European Community programme QOL 1.1.1 under contract number QKGA-1999-51330. References: 1. Szczepaniak LS et al., Am J Physiol 1999;276:E977-E989. 2. http://carbon.uab.es/mrui/. 3. Thomsen C., Magn Reson Imaging 1994;12:487-495.

Pelvis, GU 500 MRI with phased-array coil of the anal canal: normal anatomy and pathologic findings in anal fistulas and anal canal carcinoma F. Iafrate, A. Laghi, P. Paolantonio, D. Marin, M. Celestre, I. Sansoni, R. Passariello; Department of Radiological Sciences, University of Rome, Rome, ITALY. Purpose: The aim of our work was to assess the normal anatomy of the anal sphincter complex by the acquisition of high resolution T2 weighted sequences using a phased –array coil and to describe high-resolution imaging protocols for MR imaging of the anal canal with phased-array coil.We used MRI even to provide detailed anatomical description of anal canal, with correlation with anatomical and surgical table to describe different patterns of diffusion of anal canal fistulas and carcinoma. Subjects and Methods: Twenty-seven patients, 15 males and 12 females, (mean age 41.5 y.o.) underwent MRI evaluation of the pelvic region, using a superconductive 1.5 T magnet (maximum gradient strength, 25 mT/m; minimum rise time 600 microseconds, equipped with phased-array coil. High-resolution T2-weighted Turbo Spin Echo sequences (TR, 4055 ms; TE, 132 ms; matrix 390x512; in-plane resolution, 0.67x0.57 mm) were acquired on multiple axial, sagittal and coronal planes. Images were reviewed by two gastrointestinal radiologists on a dedicated workstation to identify the structures and anatomical spaces of the anal region in order to evaluate the normal anal sphincter complex and the presence of anal canal fistulas or neoplasia. Results: Optimal image quality of the anal sphincter complex was obtained in all cases providing an overall view of the anal region. Different muscular layers were observed between the upper and lower aspects of the anal canal. In the lower part of the anal canal, internal and external sphincter muscles could be observed; in the upper part, puborectal and internal sphincter muscles were depicted. Good visualization of intersphincteric space, levator ani muscle and ischioanal space was also obtained in all cases.In all normal patients optimal image quality of the anal sphincter complex is obtained. A panoramic overview of the entire muscular complex, including puborectal, levator ani, external and internal sphincter muscles is depicted. Intersphincteric space as well as ischio-rectal fossa are also celarly evaluated.The same imaging technique is used for the evaluation of anal canal fistulas or anal canal carcinoma. MR findings are classified using the Parks fistula-in-ano classification. Optimal correlation between surgical and MR findings is obtained for inter-sphincteric, trans-sphincteric and transelevator fistulas. Discussion/Conclusion: High-resolution MR images with phasedarray coil provide optimal depiction of the anal canal and the anal sphincter complex, is comfortable for the patient and easily

Pelvis, GU depicted. MRI thanks to optimal image quality was useful for staging of anal canal carcinoma and evaluation of anal canal fistulas.

501 Endorectal Magnetic Resonance Imaging (ER-MRI) in staging prostate cancer: preliminary results in 30 patients S. Montemezzi1, A. Porcaro2, A. Borsato1, P. di Benedetto1, G. Grasso1, L. Comunale2, G. Gortenuti1; 1Servizio Radiologia, OCM Borgo Trento, Verona, ITALY, 2Dipartimento di Urologia, OCM Borgo Trento, Verona, ITALY. Purpose/Introduction: ER-MRI has effectively advanced in staging prostate cancer thus allowing precise and effective therapeutic planning. At the Civil Major Hospital in Verona it is going on a prospective study which is focused on evaluating the accuracy of ER-MRI for staging prostate cancer before radical prostatectomy. Material and Methods: Herein we evaluated 30 patients of this series. The average age was 66 years (range 54 to 77) and the average preoperative total prostate-specific antigen (PSA) level was 12 ng/mL (range 2 to 54). The procedure was performed by using the Magnetom Symphony 1.5 Tesla. ER-MRI was performed 2 months after prostate transrectal biopsies. Patients were staged for having intracapsular disease when at least one of the following issues were detected: normally hyperintense prostate tissue between the tumor and capsule, clear delineation of the capsule despite broad contact between the tumor and capsule, and smooth capsular bulge. Suspicious extracapsular penetration included irregular bulge of the capsule, irregularity of the gland contour, disruption of the capsule, hypointense focal thickening or retraction of the capsule adjacent to the tumor. Histological examination was performed according to the Standford protocol. ER-MRI findings concerning prostate cancer extracapsular disease were statistically correlated with histopathologic specimens. Results: Of the 30 patients, 28 (93,33%) had stage T2 disease by ER-MRI, and 2 (6,66%) stage T3 disease by ER-MRI. Histopathology detected 27 (90%) patients as having pT2 disease and 3 (10%) patients with pT3 disease. The sensitivity, specificity, positive predictive value, and negative predictive value of ER-MRI for detecting extracapsular disease were respectively 66,66%, 100%, 100%, and 96,2%. The only patient who was understaged by ER-MRI for pT3a disease had histopathological findings showing focal extracapsular penetration of the tumor which did not involve the surgical specimen. Conclusions: In our preliminary experience, ER-MRI showed the highest specificity and positive predictive value for staging pT2 from pT3 tumors. ER-MRI findings suspicious for capsular penetration will suggest not performing nerve-sparing radical prostatectomy for the effective risk of having a positive surgical specimen A better definition of suspicious issues suggesting capsular penetration will improve ER-MRI in staging clinically localized prostate cancer.

S267

502 MR and CT Virtual Cystoscopy: protocols in comparison V. Panebianco, I. Sansoni, S. Visconti, M. Celestre, A. Laghi, D. Tancredi, R. Passariello; Radiological Sciences, Policlinico Umberto I, University of Rome, Rome, ITALY. Purpose: To create protocols for MSCT and MRI Virtual Cystoscopy, to detect and stage bladder lesions using axial CT and morphological T1-w and T2-w imaging in association with CT and MR Virtual Endoscopy. Subjects and Methods: 23 patients with bladder lesion underwent MR-examination and 15 also MSCT. Conventional Cystoscopy was performed as gold-standard. CT patient preparation and protocol included intraluminal introduction of air with Foley catheter and the following paramethers: feed-rotation 8mm/sec, rotation-time 0,5mm/sec, 120Kvp, 120mAs, slice-collimation 3mm, reconstruction 1. Before MR examinations, bladder was completely emptied and 1 cc of Furosemide was given to speed up contrast medium urinary excretion.. High resolution FSE T2-weighted sequence and postcontrast FSE or GRE T1-weighted imaging were used to improve bladder wall visualization. After Gd-DTPA injection MR-examination was completed with 3D-FLASH sequence for Virtual Endoscopy. Results: Among 21 patients underwent MRI (2 were claustrophobic), only 19 lesions were depicted at Virtual Cystoscopy, but all lesions were evident on morphological images (size 6–20 mm). Considering patients underwent also MSCT, all 15 lesions (5-20 mm) were evident on multiplanar MSCT reconstructed images and Virtual Cystoscopy. Concerning morpholocical characteristics of the lesions, we obtained the following appearance: 7 (30%) sessile, 3 (15%) peduncolated and 13 (54%) vegetating (3 ulcerated). 85% of lesions arised from bladder wall while 15% from prostate. In 15 cases we visualized anterior wall around the bladder base. Conclusion: Considering inherent advantages and limitations of CT and MR, MR-Cystoscopy has better cost-benefit ratio for patients, except for identification of smallest lesions (5mm).

503 Renal tumor's pseudocapsule. MRI appearance and diagnostic accuracy. MRI-histopathological comparative study V. J. Dombrovsky; Radiology, Rostov State Medical Univ., Rostov-on-Don, RUSSIAN FEDERATION. Purpose: This study is devoted to estimation of magnetic resonance imaging (MRI) role and accuracy in pseudocapsule (PC) detection and its safety in benign and malignant renal neoplasms. Subjects and Methods: MRI data of 634 patients with proven adenoma (n = 15), oncocytoma (n = 12), renal cell (n = 486) and transitional cell (n = 29) carcinomas, Wilms' tumor (n = 53), congenital mesoblastic nephroma (n = 2), angiomyolipoma (n = 27), lipoma (n = 3), hemangioma (5), angiofibroma (2) were retrospectively analyzed. MRI - histopathological comparisons (coefficients of correlation (r) and association (r1); p < 0,05) were created with the use of variance analysis. MRI was performed on 0,5 T unit (Tomikon S 50 Avance, "Bruker", Ettlingen, Germany). Spin-, gradient-echo sequences, including fast ones with breath-hold and fat-suppression techniques were used. Results: The PC appears as a low-intensity bend or rim around the tumor's nodule on T1- and T2- weighted images (WI). It had been

S268

Pelvis, GU

determined, that these MR signal properties depend on the PC morphological structure, containing two components – fibrous tissue and compressed renal parenchyma. MRI can detect the tumor's PC thickness difference. Usually, the PC intraparenchymal part is more expressed due to atrophic changes in surrounding renal tissue and subcapsular – feebly marked on MR images. The T2- WI spin-echo sequence (TRARE with respiratory compensation) and T1- WI gradient-echo breath-hold technique (GEFC) were more effective for PC detection in this study. The fat-suppressed T2- WI appeared to be the best in PC destruction discovering as an evidence of tumor invasion. But it was impossible to recognize the initial stage of this process. High positive correlation level between MRI and pathologic findings, concerning the presence (r1 = 0,91) and PC thickness (r = 0,85) for all renal masses was found. The same index of PC safety was high for all lesions (r1 = 0,90), except malignant neoplasms (r1 = 0,71). Conclusion: MRI has confirmed to be an accurate diagnostic tool for true pathologic status of renal tumor's PC reflection. Nephron sparing surgery based on MRI data could be indicated.

504 MRI diagnostic value in vivo renal cell carcinoma's histological structure investigation V. J. Dombrovsky; Radiology, Rostov State Medical Univ., Rostov-on-Don, RUSSIAN FEDERATION. Purpose: The prognosis accuracy of magnetic resonance imaging (MRI) data, concerning renal cell carcinoma's (RCC) structure, according to modern histological classification (WHO, 1998), was investigated in the present study. Subjects and Methods: MRI data of 671 patients, 387 men and 284 women, with a mean age of 56,1 + 4,1 years (range 18 - 82 years), with proven four main cell RCC variants were retrospectively analyzed and compared with the histopathological and immunohistochemical findings, using discriminant analysis, as well as regression and factorial analysis's of variance (r) with a 5 % level of significance (p < 0,05). MRI was performed on 0,5 T unit (Tomikon S 50 Avance, «Bruker», Ettlingen, Germany). Spin-echo (MSSE, TRARE, MYUR) and gradient-echo (GEFI, GEFC, GEDS) sequences, including fast ones with breath-hold and fat-suppression techniques were used. Results: Previously, we have found the high positive correlation level (r > 0,89; p < 0,05) between MRI and pathologic data, concerning RCC growth type, spreading, secondary alterations of tumor structure and its localization in renal parenchyma. Just the complex of these features was used for MRI prognostication of RCC cell variants. 398 (59,3 %) cases of RCC were classified by histopathological and immunohistochemical methods in respect to clear cell carcinoma, 64 (9,5 %) - to chromofilic cancer, 44 (6,5 %) - to chromophobe and 33 (4,9 %) - to collecting duct carcinomas. 132 (19,8 %) RCC cases of a mixed histological structure had been excluded from observation. MRI identified 326 (81,9 %) of 398 cases with clear cell carcinoma, 45 (70,3 %) of 64 - with chromofilic cancer, 30 (68,1 %) of 44 cases - with chromophobe and 21 (63,6 %) of 33 - with collecting duct carcinomas. The type and significance of secondary alterations (haemorrhage, necrosis), a cyst's formation, as an evidence of native tumor growth, and its localization appeared to be better predictors of RCC histological structure.

Upon statistic analysis, the positive correlation level (r = 0,71; p < 0,05) between MRI and pathologic data, concerning RCC histological variants, was found. Conclusion: MRI has confirmed to be an accurate diagnostic tool for sufficient prediction of RCC cell variants, especially clear cell carcinoma , which together with the stage of this lesion, determinates the treatment approaches and prognosis .

505 CT and MRI of renal cysts: Analysis of 19 cases G. M. Favagrossa1, F. Franco1, G. C. Mazza2, P. Caccia1; 1., 1° Servizio di Radiologia, Spedali Civili, Brescia, ITALY, 2., 2nd Dept of Radiology-Universitary Institut , Spedali Civili, Brescia, ITALY. Purpose: To evaluate 19 cases of cystic renal lesions imaged with CT and MRI, to establish the usefulness of MR exam. Subject and Methods: 19 patients, 11male and 8 female, 23-72 years old, underwent MRI after a CT exam of abdomen which discovered a complicated renal cyst. We used a singleslice helical CT and a 1.5 tesla equipe using both T1 and T2 - weighted sequences. Characterization cystic renal lesions relies mainly on the Bosniak classification which consists of four categories: benign simple cysts (cat I); minimally complicated cysts (cat II); indeterminate cystic renal massae that include cystic renal tumors ( multiloculated or not) and complex cysts; cystic renal cell carcinomas (cat IV). Results: CT showed: 12 cystic lesions (Bosniak cat II); 3 (Bosniak cat III); 3 (Bosniak cat IV). MRI discovered 5 neoplastyc cystic lesions which had been classified on Bosniak classification at CT exam : Cat II in 2 cases; Cat III in 1 case and Cat IV in 2 cases. These 5 patients underwent surgical treatment and histological exam confirmed a renal cell carcinoma in all patients. In the other 14 patients MRI showed a benigne simple cyst (5 cases, Bosniak Cat I) and in 9 patients a minimally complicated cyst ( Bosniak Cat II). Discussion: MRI is superior to CT in distinguishing a hemorragic cyst or multiloculated cystic masses. MRI is very helpful for further differential diagnosis of lesions which are discovered on CT, especially in the differentiation between complicated cysts and cystic or hypovascular renal cell carcinoma.

506 Fetal MRI: comparison of HASTE and trueFISP imaging at 1.5 T B. J. Op de Beeck1,2, R. Salgado1, F. De Ridder2, L. De Catte3, Y. Jacquemyn4, P. Parizel1; 1Radiology, University Hospital Antwerp UZA, Antwerp, BELGIUM, 2Radiology, AZ Free University Brussels VUB, Brussels, BELGIUM, 3Gynecology, AZ Free University Brussels VUB, Brussels, BELGIUM, 4Gynecology, University Hospital Antwerp UZA, Antwerp, BELGIUM. Purpose: To evaluate the image quality of two ultra-fast T2weighted sequences (HASTE and trueFISP) in prenatal MRI. Subjects and Methods: We included 23 patients with a suspicion of fetal abnormalities on ultrasound. Imaging was performed on Magnetom Vision and Sonata 1.5T scanners (Siemens, Erlangen, Germany). HASTE (Half Fourier Single Shot Spin Echo) and trueFISP sequences were used in the 3 anatomical planes. Diagnostic confirmation was obtained with fetal ultrasound and post-partum imaging.

Spine Results: The majority of indications were diseases of the brain (18) going from simple hydrocephalus to bleeding, holoprosencephaly, agenesis of the corpus callosum, arachnoid cyst and Dandy Walker variant. Other indications were meconium peritonitis, campomelic dysplasia, pulmonal hypoplasia, diaphragm hernia and sacrococcigeal teratoma. Good correlation was obtained with the fetal ultrasound findings. TrueFISP images were slightly superior to HASTE images for neurological and non-neurological diseases. Conclusion: While ultrasound remains the standard for prenatal diagnosis and screening, MRI has become an important tool in the diagnosis and management of some types of fetal abnormalities. Prenatal MRI is particularly helpful for: 1.improving anatomic definition, 2.clarifying the diagnosis and 3.identifying associated abnormalities. In our experience, trueFISP images are slightly superior to HASTE images.

Spine 507 Low back pain disability and MR findings. Relation to health care provider E. Arana1, L. Martí-Bonmatí1, M. Vega2, E. Mollá1, S. Costa1, D. Bautista3; 1Radiology, Clinica Quirón, Valencia, SPAIN, 2Radiology, Hospital Univ. Dr. Peset, Valencia, SPAIN, 3Preventive Medicine, Hopsital Univ. Dr. Peset, Valencia, SPAIN. Purpose: To evaluate differences on lumbar spine MR imaging findings and outcome measures questionnaire between different health care providers. Material and Methods: A total of 278 patients submitted with low back pain (LBP) were studied. There were 132 men and 146 women, aged 44 ± 13.8 years. One hundred and nine patients were from the National Health System (NHS) and 169 from private practice. Patients with previous surgery, neoplasm or traumatic episodes were excluded. Every patient completed a disability questionnaire with six core items 1, providing an index of disability from 0 to 28. All patients had sagittal spin-echo T1 and fast spinecho T2, axial proton-density and MR myelography weighted images. MR images of the two most affected disc levels were read offering an MR index of affectation, from zero to 30. Results: MR index correlated only with pain interference with normal work. (Kruskal-Wallis, p=0.04). No other associations were found with the other disability variables. Patients from the NHS showed greater disability than private ones (14.3±3.7 vs. 13.3±3.2, Student t-test, p=0.01) and higher MR imaging index (MannWhitney U test, p=0.03). The disability index was not statistically associated with any particular disc level. Conclusion: In patients with LBP, MR imaging does not correlate with physical self-report disability 2 with the exception of working interference. Differences are found between private and public patients, the latter being more physically affected. References: 1. Deyo RA, Battie M, Beurskens AJ et al. Outcome measures for low back pain research. A proposal for standardized use.Spine 1998 Sep 15;23:2003-13 2. Hollingworth W, Dixon AK, Todd CJ et al. Self reported health status and magnetic resonance imaging findings in patients with low back pain. Eur Spine J. 1998;7:369-75.

S269

508 1H MRS for classification of the state of vertebra body V. Rogozhyn1, Z. Rozhkova1, E. Pedachenko2, S. Kushchaev2, A. Garmish2; 1Mri, Radiological Center of the Academy of Medical Sciences of Ukraine, Kyiv, UKRAINE, 2Spinal Neurosurgery, Neurosurgery Institute of the Academy of Medical Sciences of Ukraine, Kyiv, UKRAINE. Purpose: The aim of our study is to describe the vertebral body state on the basis of 1H MRS data of the vertebra in the norm and under pathologies. Subjects and Methods: Three groups of subjects are examined by MRI and 1H MRS using 1.5 T Magnetom Vision (SIEMENS). The 1st group (G1) includes 30 healthy volunteers (23 - 64 y). The 2nd group (G2) consists of 23 female patients (42 - 73 y) with vertebral osteoporosis. The 3rd group (G3) consists of 35 patients (36 - 67 y) with vertebral hemangioma. The spectra are recorded with STEAM SVS:TR/TE=5000/20 ms, VOI=1cm3, NS=32; and 2DCSI:TR/TE=5000/20ms, PN=16x16, NS=1. In the G1 VOI was placed in L2 vertebral body. In the G2 and G3 VOI was placed in the vertebral body, identified on MR imaging as abnormal. Results: We consider the signals from water (4.7 ppm) and the following lipid fractions: -CH3 (1.0 - 1.1 ppm), methylene protons (CH2)n (1.5 - 1.6 ppm), CH2, α to double bond (CH-CH2-(CH2)n) (2.3 ppm), CH2, α to carboxyl (CH-CH2-(CH2)n) (2.5 ppm), and olefinic protons H-C=C-H (5.4 ppm). We define two observed quantities: the compound content AM as the peak amplitude and the compound concentration CM as the ratio of the peak amplitude AM to the sum of all the peak amplitudes S: CM=AM/S. The sum S=ALip+AWat we call the total content. We consider the lipid content and concentration as functions of S and analyse how these functions change with the VOI location, and state and age of different subjects. We describe the state of vertebra in each VOI by the pair T* = {AWat,ALip}, where AWat and ALip are the peak amplitudes of the signals from water and lipids. The configuration 1* ={AWat>ALip} is observed in the spectra of young volunteers, in the spectra of patients with hemangioma and in the spectra, overlying the intact tissue of vertebrae of those patients. The configuration 2* = {AWat
509 The role of Opposed Phase Gradient Echo MR imaging in differentiating between benign and malignant vertebral lesions S. S. A. Lingawi1, Y. Ragab2; 1Radiology, King Abdulaziz University, Jeddah, SAUDI ARABIA, 2Radiology, Dr. Erfan & Bagedo Hospital, Jeddah, SAUDI ARABIA. Introduction: To determine if Opposed Phase Gradient-Echo MRI can help predict the likelihood of neoplastic or nonneoplastic lesions in vertebral bone marrow. Subjects and Methods: Twenty consecutive patients with clinically suspected vertebral bone marrow lesions and twenty non-oncological patients underwent MR imaging using Spoiled Gradient-Echo sequences with identical parameters except for the

S270 Echo Time (TE), which was chosen to image fat and water in phase or out of phase. The signal intensity of the bone marrow in both groups was visually assessed on the in-phase and out-of- phase images. Two reviewers who were unaware of the patients’ identities and clinical histories assessed the images independently. The reviewers assessed decreased marrow signal intensity relative to control sites on the out-of-phase and in- phase images. The MR findings were compared with the routine bone scintigraphy. The patients had either established diagnoses or determination of benignity based on stability of findings at 1 year follow up. Results: Opposed phase gradient-echo MR imaging sequences were able to detect all metastatic lesions that were demonstrated on the nuclear medicine bone scintigraphy. These lesions appear hyperintense relative to normal vertebrae on the in-phase sequence and fail to suppress on the out-of-phase sequence. New sites of suspected vertebral metastasis were also detected in two patients without corresponding abnormalities on the bone scintigraphy. Conclusion: Opposed phase gradient-echo MR imaging sequences are recommended for the detection of skeletal metastases of the spine.

510 MRI of the spine and 99Tc MIBI scintigraphy in the diagnosis of multiple myeloma J. Nekula1, M. Myslivecek2, J. Bacovský3, D. Horák1; 1Radiology, Faculty Hospital, Olomouc, CZECH REPUBLIC, 2Nuclear Medicine, Faculty Hospital, Olomouc, CZECH REPUBLIC, 3Internal Medicine, Faculty Hospital, Olomouc, CZECH REPUBLIC. Introduction: We have examined 138 multiple myeloma (MM) patients using MRI of the spine and 145 patients using 99Tc MIBI, however, with large time intervals between the two examinations. In the present work, we have verified the validity of both methods in a group of 42 patients, where the temporal distance was not greater than 14 days. Subjects and Methods: We have examined 42 patients, 28 males, 14 females, age range 19 - 79 years, median age 61 yrs, with these diagnoses: 11 monoclonal gammapathies, 31 multiple myeloma Durie - Salmon I - III, including 15 patients after therapy. Examination of Th - LS spine was conducted on a Siemens Maestro Symphony 1.5 T, using T1, T2 and STIR sequences and, in half of the cases, GRE in the sagittal and transverse plane. Classification of findings: normal; diffuse, local or mixed form of MM; epidural or paraspinal propagation; vertebral compression. Whole-body scintigraphy in anterior and posterior projection, 7040 MBq 99Tc-MIBI, scintillation camera with two detectors. Classification of findings: type N: normal distribution, type D: diffuse accumulation in the marrow of the whole body, type F: focal accumulation, type D+F: combined diffuse and focal accumulation. Results: Pathological bone marrow changes were detected in 94% of MRI and 95% of scintigraphy exams (including 4 extraaxial foci), almost always in agreement with clinical findings, epidural mass in 22% of MRIs, vertebral compression in 49%. Scintigraphy was twice positive in patients with falsely negative laboratory and MRI findings. When following therapy, normal scintigraphy was in 89% in agreement with the clinical picture but MRI only in 22%. Conclusion: In the diagnosis of new forms of MM, detection of pathological bone marrow changes with MRI and 99Tc MIBI scintigraphy shows the same sensitivity. MRI detects epidural

Spine propagation, which substantially influences the therapeutic procedure; on the other hand, MIBI scintigraphy is more sensitive in therapy evaluation, as pathological changes on MRI have a relatively long latency.

511 The role of Diffusion-weighted MR imaging in pyogenic and tuberculous spondylodiscitis C. Calli1, O. Kitis1, T. Yurtseven2, T. Yilmaz1, N. Yunten1; 1Radiology, Ege University Medical School, Izmir, TURKEY, 2Neurosurgery, Ege University Medical School, Izmir, TURKEY. Purpose: To document the diffusion weighted MR imaging findings in patients with pyogenic and tuberculous spondylodiscitis. Subjects and Methods: 32 patients with spondylodiscitis diagnosed on standard MRI and proved either by surgery or laboratory findings were included in the study (pyogenic=23, tuberculous=9). The ages of the patients ranged from 8 to 60 years-old. On MRI, sagittal and axial turbo spin echo T1W, T2W, and post-gadolinium images were obtained. Diffusion weighted imaging (DWI) was performed in all patients at the sagittal plane using PSIF (reverse FISP) sequence on a 1.5T MR unit. The signal intensities of the infected disc, vertebrae, epidural and paravertebral infectious masses on DWI were evaluated. Results: On postcontrast MRI all of the infectious lesions showed enhancement in the intervertebral discs, affected vertebrae and the paravertebral /epidural infectious mass consistent with spondylodiscitis. In 21 (tuberculous=3, pyogenic=18) of 32 patients with acute spondylodiscitis, DWI revealed hyperintensity in the infected disc and the vertebrae. 15 of 21 patients had paravertebral/epidural infectious mass which were all hyperintense on DWI. However, in 11 patients with longstanding spondylodiscitis (tuberculous=6, pyogenic=5), DWI showed iso-hypointensity in the vertebrae and the disc, whereas accompanying paravertebral-epidural infections (in 8 patients) disclosed hyperintensity. Conclusion: DWI seems to be effective in the differentiation of acute and longstanding spondylodiscitis. Tuberculous and pyogenic spondylodiscitis have similar DWI findings in the acute stage. Moreover, they both have similar findings in the chronic stage; and DWI of the spine seems to be ineffective in the differentiation of tuberculous and pyogenic spondylodiscitis.

512 The evaluation of spinal dysraphism with CISS sequence on MRI O. Kitis1, C. Calli1, T. Yurtseven2, T. Yilmaz1, N. R. Sener1, N. Yunten1; 1Radiology, Ege University Medical School, Izmir, TURKEY, 2Neurosurgery, Ege University Medical School, Izmir, TURKEY. Purpose: The aim of this study is to investigate the role of CISS (constructive interference in steady state)sequence on MRI in the cases of spinal dysraphism. Subjects and Methods: 26 patients with spinal dysraphism (9 males, 17 females; ages ranging from 3 months to 30 years)underwent MRI of the spine with turbo spin-echo T1W, T2W and 3D CISS sequences. Orthogonal, oblique and curved multiplanar images were reconstructed from the base sagittal 3D CISS sequence images. Results: The classification of spinal dysraphism was assessed on conventional MRI sequences in all the patients (type I split cord

Head and Neck malformation=6, type II split cord malformation=8, myelomeningocoeles=5, meningocoeles=4, spinal lipoma=2, tight filum terminale=2, syringomyelic cavities=7, fibrolipoma of filum terminale=5, vertebral segmentation anomalies=7). Small syringomyelic cavities, septations, and fine dysplastic neural tissues in myelomeningocoeles and attachment of the filum terminale which were not seen on conventional images were well demonstrated on CISS sequence. In addition, more anatomical detail with high T2 resolution and multiplanar reconstruction (MPR)capability were achieved with this sequence. Conclusion: CISS sequence imaging is useful in the definition of complex abnormalities seen in spinal dysraphism which provides additional information for surgical planning.

513 Closed spinal dysraphism in adult patients D. S. Ilic-Jankovic1, D. S. T. V. Ilic-Jankovic2; 1Department of radiology Nis, Clinical Center Nis, Nis, YUGOSLAVIA, 2Department of Radiology Nis, Clinical Center Nis, Nis, YUGOSLAVIA. Purpose: The aim of this study is to present the radiological findings and results of neurosurgical intervention in two adult patients with closed spinal dysraphisms. Patients and Methods: Review of two mail patients (26 and 32 years old) presented with symptoms and signs of myelopathy. Both of them underwent clinical, x-ray, myelography, computed tomography (CT) myelography and MRI workup. Results: In both of the patients the spinal canal anomaly and spina bifida were present. The first patient (26 y.), beside this had deep extension of subcutaneos lipoma thought the spina bifida and dural deficiency to insert directly into the defect on the dorsal half of the cord. Ballooning of the subarachnoidal space to form meningocele was in diameter LL 58mm, AP 4omm and in length 14omm. Careful analysis of imaging findings indicated lipomyeloschisis ( lipomyelomeningocele). The second mail patient (32 y.) clinically had sphincter dephiciency (bladder and rectum), with other signs of myelopathy. Radiological findings indicate the large anterior sacral meningocele with sacral dysgenesis. Surgery was proposed to both of them when they were symptomatic and with progressive symptoms. After surgery patients condition improved and stabilized. Conclusion: The prognosis of skin-covered spinal dysraphysms is very good if appropriate diagnosis and care are provided early. Neonatal resection of these congenital anomalies is necessary to prevent tethering of the spinal cord that might lead to irreversible neurological damage if not released. Prognosis in adult patients without irreversible neurological damage using appropriate neuroimaging and surgical procedures is also satisfactory.

S271

Head and Neck 514 Preoperative virtual endoscopy of the cerebellopontine angle: correlation with surgical findings V. Nowé1, D. De Ridder2, P. Van de Heyning3, F. L. Wuyts3, A. M. De Schepper1, P. M. Parizel1; 1Department of Radiology, Universitair Ziekenhuis Antwerpen (University of Antwerp), Antwerp, BELGIUM, 2Department of Neurosurgery, Universitair Ziekenhuis Antwerpen (University of Antwerp), Antwerp, BELGIUM, 3Department of ENT, Universitair Ziekenhuis Antwerpen (University of Antwerp), Antwerp, BELGIUM. Purpose: To determine the clinical value of virtual endoscopy of the cerebellopontine angle (CPA) in preoperative planning and to correlate virtual endoscopic findings with peroperative results. Subjects and Methods: Methods: Thin section (0.6 mm), high resolution (100 mm FOV, 256 matrix) heavily T2-weighted images were obtained with a gradient echo sequence (constructive interference in steady state (CISS)) through the CPA. Fly-through software on an independent workstation (Leonardo, Siemens) provided virtual endoscopy images of the CPA. Subjects: Ten patients were examined prior to undergoing a neurosurgical procedure in the posterior fossa via retrosigmoidal approach. Peroperative images were compared with preoperative virtual endoscopy findings, and correlated for anatomic accuracy with regard to the course and shape of the 7th and 8th cranial nerves, blood vessels and internal auditory canal. All images were compared independently by two neuroradiologists, a neurosurgeon and an ENTsurgeon. Results: Excellent anatomic correlations were found between preoperative virtual endoscopy images and peroperative findings. The course of nerves and blood vessels in the CPA was accurately depicted by virtual endoscopy. In patients with a narrow CPA, anatomic landmarks were not as clearly visible. Conclusion: Virtual endoscopy of the CPA provides a non-invasive view of the anatomical relationships between nerves and blood vessels, and can be an important tool in planning a surgical procedure. The best correlations were found in patients with a wide CPA due to atrophy, or tumor.

515 Virtual endoscopy of the cerebellopontine angle in patients with tinnitus V. Nowé1, D. De Ridder2, P. Van de Heyning3, X. L. Wang1, A. M. De Schepper1, P. M. Parizel1; 1Department of Radiology, Universitair Ziekenhuis Antwerpen (University of Antwerp), Antwerp, BELGIUM, 2Department of Neurosurgery, Universitair Ziekenhuis Antwerpen (University of Antwerp), Antwerp, BELGIUM, 3Department of ENT, Universitair Ziekenhuis Antwerpen (University of Antwerp), Antwerp, BELGIUM. Purpose: To investigate by means of MR imaging of the cerebellopontine angle (CPA) whether a correlation can be found between the location of microvascular compressions on the vestibulocochlear nerve and three different clinical subtypes of tinnitus. To show the accuracy of virtual endoscopy imaging to evaluate the course of nerves and blood vessels in the CPA.

S272

Musculoskeletal

Subjects and Methods: Methods: Thin section (0.6 mm), high resolution (100 mm FOV, 256 x 256 matrix) heavily T2-weighted images were obtained with a gradient echo sequence (constructive interference in steady state (CISS)) through the CPA. Fly-through software on an independent workstation (Leonardo, Siemens) provided virtual endoscopy images of the CPA. Subjects: Patients were referred from the tinnitus outpatient clinic for virtual endoscopy of the CPA. All patients with known causes of tinnitus were excluded. In 47 patients the clinical subtype of tinnitus was correlated with the imaging findings: 1) non-pulsatile tinnitus high pitch; 2) non-pulsatile tinnitus low pitch; 3) pulsatile tinnitus arterial. Results: We found a good correlation between the clinical presentation of tinnitus and the images on MR. In patients with arterial pulsatile tinnitus, we identified a significant higher number of vascular loops in the internal auditory canal compared to patients with non-pulsatile tinnitus. In patients with non-pulsatile tinnitus, we found an opposite distribution of the low and high frequencies along the cisternal segment of the cochlear nerve. These findings are suggestive for a tonotopic structure of the cochlear nerve. Conclusion: Virtual endoscopy of the CPA provides a non-invasive view of the anatomical relationship between nerves and blood vessels, and can be of use to demonstrate neurovascular conflicts. Our preliminary findings indicate that, in some patients, the location of the blood vessel impinging on the vestibulocochlear nerve can be correlated with the clinical subtype of tinnitus.

516 High resolution MRI in evaluation of normal and pathologic eye and orbita B. Guney, H. Altay, C. Cinar, R. Savas, H. Alper; Radiology, Ege University, Izmir, TURKEY. Introduction: The purpose of this study was to demonstrate the small intraorbital anatomic structures (particularly septum orbitale) by using surface coil ( SC ) and to evaluate diagnostic superiority of high resolution orbital MR images, obtained with surface coil to orbital MR images obtained with head coil ( HC ) by comparing them. Subject and Methods: 52 cases were examined with bilateral orbital MR images by using HC. Additionally, in this 52 cases, the pathologic eye or orbita was examined with unilateral high resolution orbital MRI by using SC. Investigation was done by using 1.5 Tesla MRI system ( Siemens, Magnetom Vision, Erlangen, Germany). The MR images obtained by using HC and SC were compared in the evaluation of the small anatomic structures such as the aponeurosis of levator palpebrae, the superior tarsal ( Muller’s) muscle and the superior orbital ligament which are adjacent to septum orbitale and detecting intraorbital lesions. Results: The high resolution MR images by using SC demonstrated the orbital anatomic structures successfully. In 45 ( 86.5% ) cases, the septum orbitale which is difficult to be demonstrated by the other techniques shown successfully and the differentiation of levator palpebrae aponeurosis and Muller’s muscle was achieved in 47 ( 90% ) cases. Additionally 2 small lesions (choroidal hemangioma and macular degeneration) which could not be detected by using standard HC were demonstrated by SC. In 4 ( 8% ) cases, the images obtained by using SC were non-diagnostic due to movement artifacts. Discussion: The high resolution MRI by using SC is a technique which enables to demonstrate small intraorbital anatomic structures

successfully in minutes. It is superior to all the other techniques to demonstrate the palpebral layers, the septum orbitale, the levator muscles and the intraorbital compartments. The most important disadvantage is movement artifacts. It is crucial to cooperate with the patient to minimize movement artifacts and increase the quality of the examination.

517 The surface coil MR imaging and MR sialography findings of the parotid gland involvement in primary sjogren’s syndrome R. Savas, E. Erginoz, H. Alper; Radiology, Ege University, Izmir, TURKEY. Purpose: To describe the surface coil MR imaging findings of parotid gland involvement in primary Sjogrens’s syndrome and to compare the findings of surface coil and head coil MR imaging. Subject and Methods: MR imaging and MR sialography of both parotid glands were performed in 30 patients (28 women, 2 men; age range, 21-66 years) who met the diagnostic criteria for primary Sjogren’s syndrome.All MR images were obtained by using a 1.5 T superconducting system (Magnetom Vision; Siemens, Erlangen, Germany). A 13- cm- diameter circular surface coil were used to obtain high spatial resolution after head coil MR imaging performed. Surface coil MR imaging and MR sialographic findings were compared with head coil MR imaging. Results: The staging of parotid gland involvement was determined by MR findings. 7 of the 30 patients had abnormalities on MR imaging and MR sialography with head coil. However, we showed abnormal findings in nine patients using surface coil MRI. Two patients with normal findings on head coil MRI had stage I disease. Surface coil images showed better detail of parotid gland (parenchyma and intraglandular duct) than conventional MR imaging. Conclusion: Surface coil MR imaging and sialography appears to have a higher diagnostic accuracy in the evaluation of the stage of parotid involvement in patients with Sjogren’s syndrome.

Musculoskeletal 518 New MR muscle screening technique in patients with neuromuscular diseases Ö. Özsarlak, P. M. Parizel, J. W. Van Goethem, A. M. Schepper; Department of Radiology, University Hospital Antwerp, Antwerp, BELGIUM. Purpose: The aim of this study is twofold: first to introduce a new MR examination technique in screening of muscles in patients with neuromuscular diseases. Second to present the role of MR imaging in the diagnosis and differential diagnosis of neuromuscular diseases. Methods and Materials: CT examination technique of muscles introduced by Baert in the early 80’s, modified and applied for MR screening of muscles in patients with neuromuscular disease. Fifty-four individual muscles at neck, shoulder, abdomen, pelvic girdle, upper and lower leg levels studied in more than 40 patients. T1-, T2-weighted images with and without fat suppression were performed. Muscle density or signal intensity changes, volume and the symmetry of involvement were quantified. The study group consists of patients with subgroups of muscular dystrophies, varous

Musculoskeletal types of myopathies, motor neuron diseases and peripheral neuropathies. Results: Low density or high signal intensity (both on T1- and T2weighted images) changes (fatty infiltration), and decrease in muscle volume (atrophy) were major findings of neuromuscular diseases. There are 12 different patterns of fat involvement. For muscular dystrophy, fatty infiltration was more specific than atrophy, and the muscle involvement patterns showed high degree of symmetry. For motor neuron disease and peripheral neuropathies, the most significant finding was atrophy. Conclusion: Fatty infiltration and atrophy of muscles are major findings in neuromuscular diseases. Selective involvement of muscles, and different involvement patterns are specific for certain neuromuscular disease. We introduce a new MR examination technique to screen the muscle involvement in 6 different body levels.

519 Whole-body MRI for the detection of skeletal metastases in cancer patients and in benign pathology of bones J. Barceló1, J. C. Vilanova1, M. Villalón1, E. Riera2, A. Rubió2; 1Magnetic Resonance, Clinica Girona, Girona, SPAIN, 2Nuclear Medicine, Clinica Girona, Girona, SPAIN. Purpose: Presentation of our preliminary experience with a WBMRI technique with an automatic moving table as a screening tool for metastases in patients with cancer, and possible application also in benign pathology of bone with multiple or systemic localization. Materials and Methods: Fast WB-MRI was performed in 24 patients with histologically known malignant tumors and clinical suspicion of bone lesions. The automatically moving table was used for fast T1-SE and STIR sequences covering nearly the whole skeleton. The total time of acquisition was only 18 minutes. We compared WB-MRI and bone scintigraphy findings. Metastatic lesions were confirmed by follow-up over 6 months or biopsy. We accomplished the same protocol in 3 patients with benign pathology of bones. Results: WB-MRI was superior to bone scintigraphy in predicting lesions dignity with a sensitivity of 100% (bone scintigraphy: 78%), a specificity of 93% (scintigraphy: 67%) and an accuracy of 96% (scintigraphy: 71%). WB-MRI showed additional metastases in extra-skeletal regions of the body. Conclusion: WB-MRI with automatic moving table technique is an effective method for evaluating the entire skeleton in patients with suspected bone metastases. WB-MRI has significant higher sensitivity and specificity to scintigraphy in the detection of bones

S273

metastases, and also can detect extra-skeletal lesions. This technique is also useful to evaluate benign pathology of bones with possible multiple locations.

520 Bone marrow edema associated with injury of shoulder joint B. Drugova; Dept. of Radiodiagnostics, Na Homolce hospital, Prague 5, CZECH REPUBLIC. Purpose: Magnetic resonance imaging (MRI) has expanded in recent years and plays now important role also in the diagnosis and evaluation of acute injuries and chronic posttraumatic changes of the shoulder. The status, integrity and traumatic changes of the labrocapsular complex, rotator cuff and scapulohumeral muscles as well as the responses of the bone marrow to trauma can be evaluated. The bone marrow edema has been found in occult fractures, in bone bruises and in an inflammatory and degenerative arthropathies. It is generally accepted that knowledge about the bone marow signal abnormalities may help in diagnostics of a wide range of disorders. The purpose of this study was to evaluate an incidence of the bone marrow edema in the head of the humerus after injuries of the shoulder joint and associated structures. Subjects and Methods: During a 24 - month period, 70 patients (45 men and 25 women ; age range 19 – 77 years, mean age 43,4 years ) after injury of the shoulder were examined. MR studies were performed on a Symphony 1.5 –T unit (Siemens, Erlangen, Germany) using standardized protocol for the shoulder. The bone marrow edema area was hyperintense on STIR, T2 weighted and T1-fl2dfs images and hypointense on T1-weighted MR images. Results: The majority of examined patients exhibited multiple lesions of the soft tissues of the shoulder. The bone marrow signal abnormality indicating the bone marrow edema of the humeral head was evident in 66 % of patients while in remaining 44 % was bone marrow pattern normal. In group of patients showing the bone marrow edema was evident high incidence of the posttraumatic changes within the supraspinate tendon (91 % of cases), high incidence of the labral pathology (50 % of cases) and moderate incidence of the supraspinatus impingement syndrome and tendovaginitis of the long -head of the biceps (both 37 % of cases). Incidence of these changes was significantly lower in group of patients showing the normal signal characteristics of the humeral head bone marrow namely concerning the labral pathology (25 %) and the tendovaginitis of the long head of the biceps (17 %). Incidence of the impingement syndrome was in both groups comparable (37 % v. 42 %) Conclusion: The results of this study show relationship between the incidence of the bone marrow edema of the humeral head and posttraumatic changes of the labrum and the long head of the biceps tendon.

521 MRI and CT aspects in hemofilic gonarthropathy F. F. Birsasteanu1, D. I. Onet, Sr.1, D. V. Poenaru2, S. Motoi1, J. Patrascu2, I. Branea2; 1MRI Department, University Hospital, Timisoara, ROMANIA, 2Orthopaedic, University Hospital, Timisoara, ROMANIA. Introduction: The hallmark of hemophilia is hemorrhage into the joints. Hemophilic arthropathy evolves through 5 stages, starting as an intra-articular and periarticular edema due to acute hemorrhage

S274

Musculoskeletal

and progressing to stage 5, which consists of advanced erosion of the cartilage with loss of joint space, the presence of joint fusion, and fibrosis of the joint capsules. Subjects and Methods: We included in our study 7 male patients with severe hemophilia A, with age between 11 and 37, patients who were treated in Clinic of Orthopaedic II, County Hospital No. 1 Timisoara, between 2001 - 2002. The main aim of this study is to evaluate from imagistic perpectives knee modifications in case of hemophilia. We perform to our patients radiographs (F + P), computed tomography, MRI and lumbar and left hip osteodensitometry. Next step were to compare knee modification depending on methods of evaluation. We analysed cases using Pattersson score. All cases had a score bigger than 6. Results: Imaging Studies showed synovial hypertrophy, hemosiderin deposition, fibrosis, and damage to cartilage that progress with subchondral bone cyst formation are visualized on x-ray films and CT and occur in patients who are untreated, inadequately treated, or have recurrent joint hemorrhages. Discussion: A patient with hemophilia A or B (hereditary Factor 8 or 9 deficiency) who experiences intra-articular hemorrhage may develop a disabling arthropathy. Pettersson et al developed a radiographic score for affected joints. The presence of effective prophylaxis has reduced the severity of this complication.

522 Combined in-Vivo MRS and ex-vivo metabonomic analysis of the metabolic response of Zucker rats to a diabetogenic treatment J. Gounarides, M. Korach-Andre, S. Mandiyan, D. Laurent; Core Technology, Novartis Institute for Biomedical Research, East Hanover, NJ. Introduction: The accumulation of intramyocellular lipid (IMCL) has been suggested as a possible trigger of peripheral insulin resistance (IR). Still questions remain as to mechanisms involved. We characterized in fa/fa Zucker (ZFR) the metabolic response coinciding with dexamethasone-induced IR using both in-vivo localized 1H-MRS and ex-vivo metabonomic analysis of serum and muscle tissue. Methods: Eight ZFR were given a 11mg/kg dose of dexamethasone (dexa) at day 0, 2 and 7, while a corresponding group was dosed with saline. Serum and tibialis anterior (TA) muscles were collected at days 1 and 9, respectively. IMCL measurements were performed 3 days before the first injection of dexa and at 2 to 3 days subsequent to the 2nd injection. In-vivo 1H-MRS spectra were obtained from the TA muscle on a Bruker 3.0T/60cm instrument by using a PRESS sequence (8mm3 voxel, TR/TE=2s/18ms, 1200 averages) with water suppression. Peak areas for total creatine (tCr), EMCL and IMCL were determined using a line fitting procedure. High resolution 1H-MR spectra of serum samples and muscle extracts were acquired at either 500 or 400 MHz on Bruker spectrometers. Metabolite assignments were based on previous literature and in certain cases by spiking. Multivariate analysis of NMR data was performed using Pirouette (Infometrix Inc). Results: A marked increase (~60 %, p<0.05 vs. controls) in IMCL content was measured in the TA muscle of dexa-treated rats. This was accompanied by a ~30% increase (p<0.05) in in vivo tCr peak areas. Ex-vivo 1H-MRS analysis of TA muscle extracts indicated that dexamethasone did not alter creatine, phosphocreatine, or creatinine levels. 1H-MR analysis of the serum showed a ~115% increase in dexa-treated rats in triglyceride methyl signal area relative to controls (while glucose was increased by ~136%).

The triglyceride methylene/methyl signal intensity was increased by ~13 % in dexa-treated rats indicating longer average chain length of serum triglycerides. However no increase in muscle lipid chain length was measured. The degree of fatty acid unsaturation remained unchanged. Discussion: Our results suggest that the IMCL increase, as measured in vivo, results from an accumulation of lipid molecules rather than the synthesis of longer fatty acid chains. Ex vivo analysis showed that tCr was not affected by dexa treatment. We therefore conclude that dexa treatment increases the in vivo MR ‘visibility’ of the tCr signal. If true, this would have significant consequences for studies in which the quantification of IMCL is done relative to creatine.

523 Application of 31P NMR Spectroscopy for the early diagnosis of Sudeck Dystrophy. M. Krssak1, M. Leixnering2, M. Roden1, R. Goris3; 1Internal Medicine 3, University of Vienna, Wien, AUSTRIA, 2Experimental and Clinical Traumatology, Ludwig Boltzman Institute, Wien, AUSTRIA, 3Surgery, University Hospital, Nijmegen, NETHERLANDS. A substantial percentage of patients with Colles fracture develop Sudeck Dystrophy (RSD). RSD can lead to severe intractable pain and fibrosis of the joints. No specific diagnostic tool is available to identify this syndrome. A preliminary study detected decreased levels of phosphocreatine (PCr) in late cases of RSD. This study aims at testing of 31P NMR spectroscopy for it’s diagnostic capability in early stages of RSD. We studied 6 patients at early stages of RSD. Patients were diagnosed in emergency hospital and enrolled in a study on pathophysiology after Colles Fractures. Pulse-acquire 31P NMR spectroscopy (TR=6 sec, NS=32) of palm muscle of the RSD hand and the unaffected (CON) hand was applied to measure PCr/ATP, Pi/ATP and the pH of muscle. Patients were lying in a supine position in a 3T MR system with the muscle of interest resting on 1H/31P double tuned 35 mm surface coil fixed to a supporting table. One patient underwent an additional measurement on flexor digitorum of both forearms at rest and during exercise. The exercise consisted of 75 sec hand-grip movement (frequency 1 Hz) using ergometer with 1.4 kg resistance latex band. The PCr/ATP and Pi/ATP during exercise and recovery (300 sec) were quantified by manual integration of respective resonances from the blocks of four spectra (TR=4sec). pH was calculated using chemical shift difference between Pi and PCr resonance. PCr/ATP was lower in RSD palm muscle in 4 of 6 patients leading to the difference of the mean values (RSD: 2.74±0.24 vs. CON: 3.24±0.16, p<0.05). Pi/ATP and pH did not differ between RSD and CON. The patient who underwent an exercise test showed similar values of resting PCr/ATP (flexor digitorum, RSD 4.00 vs. CON 3.49). During exercise, the PCr/ATP in RSD decreased by 57% and did not recover till the end of study. PCr/ATP in the CON decreased only by 35% and recovered in about 90 seconds. Similar time course differences were found for Pi/ATP. Resting pH was similar in both hands (RSD: 7.07 vs. CON: 7.11). Exercise induced decrease of pH was more prolonged and pronounced in the RSD (6.76 at 146 sec of recovery vs. CON: 6.95 at 34 sec of recovery). Again, pH in the RSD did not recover till the end of study. Combination of 31P NMR spectrometry of skeletal muscle in the basal and exercise loaded state could be a promising tool for the early diagnosis of Sudeck Dystrophy.

Musculoskeletal

S275

524

525

Lack of evidence of muscle energetics impairment in macrophagic myofasciitis. A combined MRI and 31P MRS investigation D. Bendahan1, S. Guis2, J. Mattei2, J. Pellissier3, F. Nicoli4, D. Figarella-Branger3, Y. Le Fur1, G. Kaplanski5, J. Pelletier6, P. Cozzone1; 1Centre de Résonance Magnétique Biologique et

1H and 31P MR Spectroscopy of human muscles in healthy

Médicale, UMR CNRS 6612, Faculté de Médecine, Marseille, FRANCE, 2Service de Rhumatologie, Hôpital La Conception, Marseille, FRANCE, 3Service d’Anatomie-Pathologie et de Neuropathologie, Hôpital la Timone, Marseille, FRANCE, 4Service de Neurologie, Hôpital Sainte Marguerite, Marseille, FRANCE, 5Service de Médecine Interne, Hôpital La Conception, Marseille, FRANCE, 6Service de neurologie, Hôpital de la Timone, Marseille, FRANCE. Purpose: Macrophagic myofasciitis (MMF) has been recently described as a muscle disease occurring after vaccination. Histologically, inflammatory infiltrates containing aluminic inclusions have been reported while clinically MMF patients mainly complain of myalgia, chronic fatigue and intolerance to exercise. In keeping with these observations, an alteration of muscle energetics could be suspected. Objective: The aim of this study was to investigate whether MMF signs are linked to any alteration of muscle energetics using MRI and 31P MRS. Patients and Methods: Ten patients with histological signs of MMF were included in the present study and compared with 27 controls free of any muscle disease. Muscle Images were acquired on a 1.5-T MAGNETOM Vision Plus system (Siemens). Imaging was performed in the axial plane of the dominant thigh, using successively conventional T1-weighted Spin-Echo, T2-weighted Turbo Spin-echo and a double-echo turbo spin echo STIR sequences. MR spectra of forearm flexor muscles were recorded at 4.7 T (Biospec, Bruker) throughout a rest-exercise-recovery protocol. Briefly, spectra were time-averaged over 30 seconds (16 scans) and sequentially recorded during 3 min of rest, 3 min of exercice and 20 min of recovery. Exercise consisted in finger flexions performed at 0.6 Hz using a weight-pulley system. The motion amplitude of the sliding weight was measured continuously using a home-built transducer and data were processed on a personal computer using ATS software (SYSMA, France). As a standardisation procedure, metabolic changes related to PCr and pH were scaled to the power output. Relative concentrations of phosphocreatine (PCr), inorganic phosphate (Pi), adenosine triphosphate (ATP) and phosphomonoesters (PME) were obtained and intracellular pH was calculated .The rate of PCr resynthesis during the post- exercise recovery period was fitted to a first-order exponential equation and used to calculate the maximum rate of oxidative ATP synthesis. Results: No differences in the extent of pH, PCr changes, accumulation of ADP were found by 31P MRS investigation when comparing patients and controls during exercise. The initial rate of PCr recovery and the maximum aerobic capacity were also similar between the two groups. MRI investigations display unspecific signs of muscle atrophy with fatty infiltration solely in a subgroup of 3 patients suffering from nervous system involvement. Conclusion: 31P MRS and MRI investigations of MMF patients do not indicate any morphologic or metabolic anomalies suggesting the existence of a metabolic myopathy as a causative factor of muscle pain in MMF.

volunteers with positive and negative family history of hypertension P. Vyhnanovská1, A. Škoch1, P. Fendrych1, T. Pelikánová2, M. Hájek1; 1MR-unit, Dept. of Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague, CZECH REPUBLIC, 2Dept. of Diabetology, Institute for Clinical and Experimental Medicine, Prague, CZECH REPUBLIC. Purpose: Hypertension, dyslipidemia, diabetes mellitus and smoking status are synergistic risk factors for cardiovascular events. The purpose of this study was the comparison of intramyocellular and extramyocellular lipids concentrations and energetic metabolism of calf muscle, of young healthy men with positive and negative family history of hypertension. Subjects: 1H MR spectroscopy and 31P MR spectroscopy were performed on a group of 10 young men (mean age: 31.6 ± 6.0 years) with hypertension in family anamnesis (at least one of the parents) and group of 13 men (mean age: 29.5 ± 4.7 years) with negative family history of hypertension. Biochemical tests and MR examination of all subjects were performed under the same protocol after an overnight fasting (minimum 6-8 hours). Method: MR images and spectra were measured on 1.5 T Siemens Vision Imager using standard head and surface coils. T1W and T2W images were used for the localization of SVS 1H MRS. Volume of about 2ml in M.soleus was chosen for the measurement of 1H spectra (PRESS TE=135ms, TR=5000ms, AQ=64). 31P MR spectra were obtained by using surface coil which was positioned in the centre of calf. Sequence with NOE enhancement (TR=5000ms, TE=1ms) was applied. Spectra evaluation was done by Siemens Numaris software. Result: Statistical tests did not prove the difference between extra and intramyocellular lipids in both groups of subjects. Statistically 31P MRS did not show the difference in phosphorus metabolism ratio, pH and concentration of free Mg2+ (show Table). Significant differences in blood pressure, level of insulin on an empty stomach and after stimulation by glucose were found. Result of 1H and 31P MR spectroscopy CH2a /

CH3a /

CH2b*

CH3b*

PCr / Pi**

Pi / ATP** pH

[fMg2+]

patients (10) 2.9 ± 1.2 6.2 ± 5.3 7.05 ± 1.21 0.24 ± 0.04 7.07 ± 0.02 606 ± 76 controls (13) 2.4 ± 1.0 6.4 ± 4.8 7.63 ± 1.07 0.21 ± 0.03 7.06 ± 0.02 632 ± 124

* CH2a corresponds lipid-CH2-region of extramyocellular lipids at 1.47ppm; CH2b corresponds lipid-CH2-region of intramyocellular lipids at 1.29ppm; CH3a, the methyl resonance at 1.12ppm (lipid-CH3-region of extramyocellular lipids); CH3b, the methyl resonance at 0.88ppm (lipid-CH3-region of intramyocellular lipids) ** relative ratios of PCr/Pi (phosphocreatine/anorganic phosphate) and Pi/ATP (anorganic phosphate/adenosintriphosphate) Conclusion: We did not find any differences in 1H and 31P MR spectroscopy of calf muscles of men with positive and negative family history of hypertension. It is in agreement with our previous work where 31P MR spectroscopy was used for studying patients with primary juvenile hypertension [1]. This study was supported by grant CEZ:L17/98:00023001. Reference: 1. Hájek M. et al., Physiol.Res. 2002; 51:159-167

Diffusion

S276

526

Diffusion

Metabolite T1 and T2 relaxation times in human skeletal muscle at 3 Tesla M. Krssak1, V. Mlynarik2, M. Meyerspeer3, E. Moser3, M. Roden1; 1Internal Medicine 3, University of Vienna, Wien, AUSTRIA, 2Radiodiagnostics, University of Vienna, Wien, AUSTRIA, 3Medical Physics, University of Vienna, Wien, AUSTRIA. Introduction: Reasonable estimates of relaxation times are important for accurate quantification and optimizing of the measurement protocols in MR spectroscopy. Although water relaxation times of human skeletal muscle were assessed at different field strengths, in vivo data at 3 T, which becomes increasingly available for clinical routine, have not yet been reported. Methods: Experiments were performed on a 3 T whole-body MRsystem (Bruker, Germany) using a standard birdcage volume coil and STEAM localization (TR=6 s and TM=60 ms, NS=32/64, VOIsize= 1.73 cm3) in the selected volumes of soleus (SOL) and tibialis anterior (TA) muscles. Water suppression was achieved by a modified SWAMP method. For the quantification of water signals a separate spectrum (NS=4) was acquired without suppression. For the measurement of T1 relaxation times an adiabatic sech(30 ms) inversion-pulse preceded the STEAM sequence (TIR=150-1350 ms; TE=20 ms). Spectrum without inversion was acquired to assess M∞. Echo-times for the T2 relaxation times measurement varied from 10 to 200 ms. Spectra were processed using MacNuts (AcornNMR Inc., USA). Spline baseline correction was applied and the intensities of resonances of interest (IIR and ITE: water, choline/carnosine - TMA, creatine CH3 – Cr, extramyocellular lipids – EMCL, intramyocellular lipids - IMCL) were quantified using line-fitting procedures. Line splitting of resonances was adjusted to patterns according to the known B0 orientation dependence. Relaxation times were calculated with a exponential data-fittingroutine from (I∞ – IIR) vs. TIR (T1) and ITE vs. TE (T2) within Origin 5.0 (MicrocalSoftware Inc., USA). T1 measurements were performed in each muscle group in five and T2 measurements in seven young healthy sedentary volunteers. Data are presented as means ± SEM and compared using Student t-test. Results: T1 and T2 relaxation times (in ms) in human SOL and TA muscles are given in Tab. 1. T2 relaxation behaviour of Cr resonance was found to be biexponential. No differences between relaxation times were found between the muscle groups Discussion: To our knowledge, this is the first report on relaxation times of skeletal muscle metabolites at 3 T. Relaxation times of water and lipids are within the range of data reported for 1.5 T and 4 T. The biexponential spin-spin relaxation behaviour of Cr suggests a possible compartmentalization of total Cr between mobile and macromolecule bound pools, which is in agreement with results of magnetization transfer studies in animals and human brain and muscle tissue.

527 Eddy current-nulled diffusion tensor imaging by a STEAM-EPI sequence G. Steidle, F. Schick; Section on Experimental Radiology, University of Tuebingen, Tuebingen, GERMANY. Introduction: Diffusion-Tensor-Imaging (DTI) using SE-EPI in tissue with short T2 values like kidney [1] or skeletal muscle [2] suffers from a poor SNR for higher b-values. Using a stimulated echo acquisition mode (STEAM) sufficient high b-values can be obtained by increasing the mixing time (TM) rather than TE and thus not affecting T2 relaxation losses. Additional gradient waveforms are implemented in a diffusionweighted STEAM-EPI for minimizing the influence of given eddy currents as proposed in [3] but without any increase in the minimum TE or acquisition time. Methods: Gradient pulses were implemented (Fig. 1) where eddy current effects proportional to e–λ×t will vanish, if

Main source of artifacts in DTI are eddy currents with a time constant T/λ≈ ln2 where T is the time between start time of the first gradient of the pair and the begin of the read-out. For λ >> t0 the first equation simplifies to t1 = t0. For t2, λ≈ 20 ms was chosen. Eddy currents with other time constants λ between 10 and 100 ms were also drastically reduced (residual eddy current < 1%). In-vivo DTI measurements were performed in human calf musculature on a 1.5 T Siemens Sonata unit. Measurement parameters were TE = 38 ms, TM = 105 ms, TR = 4000 ms, b = 0 resp. 600 s/mm2, Matrix 64 x 64, FOV 200 mm, NEX = 4, slice thickness 8 mm. Diffusion weighting was applied in six directions in order to calculate the diffusion tensor. TE and TM were optimized to obtain a maximal SNR for the diffusion map.

T1 and T2 relaxation time in human soleus and tibialis anterior muscle at 3T T1 SOL (ms) T1TA (ms)

T2 SOL (ms) T2 TA (ms)

Water

1348 ± 51

1399 ± 16

31± 1

28 ± 1

TMA

1009 ± 107

793 ± 190

Cr – CH3 (mono)

766 ± 163

1048 ± 245

Cr – CH3 (fast)

12 ± 5

20 ± 5

Cr – CH3 (slow)

163 ± 20

199 ± 77

IMCL – CH2

354 ± 38

467 ± 42

91 ± 7

95 ± 9

EMCL – CH2

313 ± 33

446 ± 15

86 ± 5

84 ± 5

Figure 1: Diffusion-weighted STEAM-EPI sequence with additional gradient waveforms for minimizing eddy current effects. t0 = 7.5 ms, s = 2.9 ms Results: Diffusion map calculations could be performed without significant image distorsion. Fig. 2 shows an ADC map of human calf musculature.

Diffusion

S277

were aligned using the phantom calibration method [1] and a 3D computational image alignment technique [2]. The latter method involved a three-step process. Firstly, EP images with the same bmatrix were rigidly aligned together to remove patient motion. Secondly, affine transformations were used to align G2-G6 to G1. Finally, all DW-EP images were aligned to the T2-weighted acquisition. Results: Figure 2 shows FA maps generated from DW-EP images in which eddy current artefacts have been corrected using (a) water phantom derived eddy current calibration data, (b) direct warping of the distorted DW-EP images onto the undistorted T2-weighted images. These maps shows that the white matter fibres are better defined in (b) than (a), while grey matter regions and CSF have lower anisotropy in (b). Figure 2: ADC map of human calf musculature. Discussion: In a diffusion-weighted STEAM-EPI sequence gradient waveforms within TM were implemented to minimize eddy current artifacts without an increase in minimum TE and acquisition time. Optimizing TE and TM for a maximal diffusion contrast, TM will be long enough to allow positioning of the gradients for eddy current correction within TM. The implemented gradients also serve as crusher gradients to avoid unwanted coherences. References: 1. Ries M et al. [2001] J.Magn.Reson.Imaging 14:42-49 2. Sinha U et al. [2002] J.Magn.Reson.Imaging 15:87-95 3. Alexander AL et al. [1997] Magn.Reson.Med. 38:1016-1021

528 Comparison of methods for removing eddy current artefacts in brain diffusion tensor MR imaging M. E. Bastin1, P. A. Armitage2, T. Carpenter2; 1Medical and Radiological Sciences (Medical Physics), University of Edinburgh, Edinburgh, UNITED KINGDOM, 2Clinical Neurosciences, University of Edinburgh, Edinburgh, UNITED KINGDOM. Purpose/Introduction: DT-MRI tractography using single-shot EPI requires the acquisition of DW images that are free from eddy current artefacts. Such geometric distortions can be corrected using distortion parameters derived from phantom calibration experiments [1]. However, this method is time consuming and only practical for axial slices. Here, this approach is compared with a post-processing method that warps the DW-EP images directly onto the baseline T2-weighted EP images. Background: Eddy current distortions cause translation (T), shearing (S) and scaling (M) of the image in the phase encode direction. For acquisition schemes in which data is collected from 15-20 noncontiguous slices, M, T and S derived from calibration experiments vary slowly over time [1]. This means that calibration scans need not be performed after every patient examination. However, in isotropic voxel acquisition schemes, where 40-60 slices are collected, eddy current distortions become increasingly significant. This has the effect of reducing the utility of phantom calibration data, since the temporal stability of M, T and S decreases as the number of slices per TR period increases. Methods: A volunteer underwent whole-brain near-isotropic voxel DT-MRI on a GE 1.5 T scanner. Sets of axial DW-EP images (b = 0 and 1000 s/mm2) were collected with diffusion gradients applied along six non-collinear directions (G1 to G6). The DT-MRI data

Discussion: These results demonstrate the effectiveness of direct warping of DW and T2-weighted EP images. This post-processing method, while computationally intensive, produces better quality DT-MRI parametric maps than the phantom calibration approach. References: 1. Bastin ME, Armitage PA. MRI 2000;18:681-687. 2. Jenkinson M, Smith S. Medical Image Analysis 2001; 5:143-156.

529 Deriving perfusional information from diffusion tensor imaging data C. Kiefer1, L. Remonda1, H. Mattle2, K. Nedeltchev2, J. Slotboom1, H. Oswald3, K. O. Loevblad4, G. Schroth1; 1Neuroradiology, University Hospital Bern, Bern, SWITZERLAND, 2Neurology, University Hospital Bern, Bern, SWITZERLAND, 3T-Systems, T-Systems, Bern, SWITZERLAND, 4Neuroradiology, HUG, Geneva, SWITZERLAND. Purpose: In magnetic resonance imaging the intravoxel incoherent motion technique [1] measures the statistical phase effects of the velocity distribution of blood flowing within the capillary network. A model that fits the measured PGSE signal in a better way has been suggested by Henkelman et al. [2] and Ahn et al. [3]. They favorized a self-similarity approach with excellent results. We want to take up this interesting field of study and discuss two approaches for an alternative derivation of expressions for the perfusion from diffusion tensor data. Subjects and Methods: Both techniques proposed is in common that a multi b-value (b=0,50,100,1000,1200 s/mm2) diffusion tensor imaging sequence has been performed with field gradients switched along six noncolinear directions (TA = 5min). The first approach addressed here analyses the usage of a multiresolution

S278

Diffusion

fuzzy clustering algorithm for the classification of the S(b)-curves. The second technique uses a combined fit model: S = fp Sp + (1 - fp) Sd where Sd = exp( - b D) and SP is a term proposed by Ahn et al. [3] that gives an expression for the signal attenuation due to the blood flow within an fractal capillary network (indices indicate p (perfusion), d (diffusion)). Results: Six patients with an acute onset of a hemispheric stroke has been examined with the new DTI / first bolus tracking stroke protocol within a time window of six hours. These patients fullfilled the following criteria: minimal motion, pronounced areas with low perfusion but unaffected ADC and T2. The fuzzy clustering and the combined fit technique show a high mean correlation between the classified membership matrices and the TTP maps and between the velocity distribution maps and the CBF maps respectively. Conclusion: The proposed techniques allow to determine three clnically relevant informations at once: the ADC (trace of the tensor), the aniotropy maps (fiber tracks) and the perfusional disturbances. References: 1. LeBihan D, Breton E, Lallemand D, et al. Radiology 1986;161:401-407. 2. Henkelman RM, Neil JJ, Xiang Q. Mag Reson Med 1994; 32:464-469. 3. Ahn CB, Lee SY, Nalcioglu O, Cho ZH. Med Phys 1987;14:43.

530 Simulating neuronal fiber bundles for DT-MRI tracktography A. Leemans; Physics, RUCA, Antwerp, BELGIUM. Introduction: Diffusion Tensor MRI (DT-MRI) is a relatively new imaging technique that measures anisotropic diffusion of molecules in biological tissue [1, 2]. This anisotropy reflects the presence of spatially oriented structures and, therefore, allows to study the architecture of axonal fibers in the central nervous system. Several methods have been developed to reconstruct these White Matter (WM) fiber pathways [3]. However, validation of these fiber tracking algorithms remains unsatisfactory due to the lack of a gold standard for comparison. In this work, a DT-MRI phantom is presented that simulates the physical properties of a WM fiber pathway in a natural way. This allows to evaluate the numerous criteria that characterize a tractography algorithm, quantitatively. Methods: A realistic WM fiber tract should exhibit a natural curvature and should have a certain extent. Also, a smooth transition between the WM fiber bundle and the surrounding biological tissue should be employed. To compute a DT-MRI phantom that takes these aspects into account, we developed a mathematical framework. First, a set of points is generated that defines the position and curvature of the WM fiber. A piecewise continuous 3D space curve is then associated with these points, which represents the center of the fiber tract. Next, this curve is convolved with a model kernel to obtain the physical extent. The gradient of the resulting convolution is computed, which enables us to calculate the structure tensor. Finally, the first eigenvector of the tract diffusion tensor field is determined via the eigenvalue decomposition of this structure tensor. Results: An example of the simulated DT-MRI phantom is given in the figure. The color-code reflects the degree of anisotropic diffusion and the corresponding cones represent the first eigenvector of the diffusion tensor field at that position. Conclusions: We developed a DT-MRI phantom that simulates the

physical properties of a WM fiber pathway. The analytical form of this phantom allows a smooth transition between the WM fiber and its surrounding tissue, resulting in a more realistic WM fiber representation. In addition, the proposed phantom can be easily extended to more complex configurations. Acknowledgements: Funding was provided by the I.W.T. and the F.W.O., Belgium. References: 1. Basser et al, Biophysical Journal 66 (1994), pp. 259-267; 2. Le Bihan et al, Journal of Magnetic Resonance Imaging 13 (2001), pp. 534-546; 3. Mori et al, NMR in Biomedicine 15 (2002), pp. 468-480.

531 Optimisation of the diffusion contrast in spinal cord. Simulation and in vivo animal study W. P. Weglarz1, K. Majcher1, T. Banasik1, P. Brzegowy2, A. Jasinski1; 1Nuclear Radiospectroscopy, Institute of Nuclear Physics, Kraków, POLAND, 2Medical College, Jagiellonian University, Kraków, POLAND. Introduction: With use of high diffusion gradients (b above ~2000 s/mm2) the non-exponential diffusion decay was clearly demonstrated in brain and spinal cord in vitro and in vivo [1-5]. The twoexponential function was often used for approximation of signal amplitude dependence on b. The aim of this work was to optimize visualisation of the two-component anisotropic diffusion when limited number of b-values is available. Subjects and Methods: The male Wister rats were anaesthetised using halothane in oxygene. MR measurements were performed on the 4.7 T research tomograph equipped with Maran DRX (Resonance Instruments Ltd.) console and the home built gradient coils and the surface coils. ECG and breath triggered spin echo (SE) sequence with diffusion gradients were used for b-values up to 5000 s/mm2. Data were analysed using home developed software under IDL (Research Systems Inc.) environment. Results: The set of diffusion weighted MR images (128x128) of the spinal cord as a function of b-values were obtained. Dependence of signal amplitudes on b-values for directions perpendicular and parallel to the spinal cord axis for white matter and gray matter were analysed in terms of two components. Parameters of the two-exponential decay were similar to these published in [5]. The optimal diffusion contrast for diffusion gradient oriented perpendicularly to the spinal cord axis was achieved for b above ~2500 s/mm2, while for diffusion gradients parallel to the spinal cord axis better contrast was achieved for b below ~1500 s/mm2. Simulations of the influence of the b-values number and range limitations on the parameters of the two-component decay were performed and predictions were tested on experimental results. The protocol for optimal choice of the limited number of b-values was proposed.

Cells, Extracts, Fluids Conclusion: Reasonable visualisation of the two-exponential diffusion decay in nervous tissue in the case of limited number of data requires the appropriate choice of b-value. Acknowledgements: This work was supported by the State Committee for Scientific Research (KBN) of Poland under grant No: 2 P03B 102 18. 1. Niendorf T, Dijkhuizen RM, Norris DG, van Lookeren Campagne M, Nicolay K [1996] Magn. Reson. Med. 36: 847-857 2. Assaf Y, Cohen Y [2000] Magn.Reson.Med. 43:191-199 3. Inglis BA, Bossart EL, Buckley DL, Wirth III ED, Mareci TH [2001] Magn.Reson.Med. 45:580-587 4. Weglarz WP, Adamek D, Markiewicz J, Skórka T, Brzegowy P, Jasinski A [2003] Solid State NMR, accepted. 5. Elshafiey I, Bilgen M, He R, Narajana PA [2002] Magn. Reson. Imag. 20:243-247

S279

have revealed numerous iron particles in the cytoplasm of hES. On T2- weighted images, the labelled cells have shown well-defined hyopintense areas at the site of injection in anterior LV wall. Conclusions: It is feasible to magnetically label and visualise hES. MR visualisation of magnetically labelled hES may be a valuable tool for in vivo tracking of hES.

Cells, Extracts, Fluids 532 MR imaging of magnetically labelled human embryonic stem cells T. Tallheden1, M. Lorentzon2, O. Rakotonirainy2, B. Soussi2, F. Waagstein2, A. Jeppsson3, A. Lindahl1, E. Omerovic2; 1Department of Clinical chemistry and Transfusion medicine, Sahlgrenska University Hospital, Gothenburg, SWEDEN, 2Wallenberg laboratory, Sahlgrenska University Hospital, Gothenburg, SWEDEN, 3Department of Cardiothoracic Surgery, Sahlgrenska University Hospital, Gothenburg, SWEDEN. Introduction: Human embryonic stem cells have emerged as a potentially new therapeutic approach for treatment of heart and other diseases applying the concept of regenerative medicine. A method for in vivo visualisation and tracking of transplanted hES would increase our understanding of in vivo hES behaviour in both experimental and clinical settings. The aim of this study was to evaluate the feasibility of magnetic labelling and visualisation of hES with magnetic resonance imaging. Subjects and Methods: Human ES were established and expanded according to standard procedures. After expansion, the cells were further cultured in feeder-free culture conditions and magnetically labelled by addition of SPIO agent dextran-coated Ferrum-oxide particles (Endorem®, Gothia Medical) to the culture medium. Accumulation of SPIO in hES was assessed by Prussian Blue staining and electron microscopy. Viability of the cells was assessed by trypan-blue. For in vitro imaging, the labelled and unlabelled hES were placed in a standard 1 cm probe embedded in the agarose gel to reduce the susceptibility effects. The hES were then injected (10 µl) into excised mouse hearts with injection placed within anterior left ventricular (LV) wall. MRI experiments were performed on a Bruker Avance DMX 500 with a standard vertical magnet (52 mm diameter) operating at 11.75 T. The microimaging probehead was equipped with a shielded gradient system and a radio frequency (RF) insert. A saddled RF coil tuned to 500 MHz for proton was used for transmission and detection. A multi-slice multi spin-echo imaging pulse sequence consisting of 32 echoes and an echo spacing of 6 ms was used to obtain T2-weighted images. Results: hES appeared to be unaffected by magnetic labelling and maintained their ability to proliferate in the culture. No agent for membrane permeabilisation was needed for facilitation of intracellular SPIO accumulation. Prussian Blue and electron microscopy

533 Mobile lipids and glucose metabolism detection in Hut 78 cells by 1H and 13C-edited 1H MRS E. Iorio1, C. Testa1,2, C. Casieri3, F. De Luca2, F. Spadaro1, C. Ramoni1, E. Lococo4, R. Carnevale4, L. Lenti4, R. Strom5, F. Podo1; 1Laboratory of Cell Biology, Istituto Superiore di Sanità, Rome, ITALY, 2INFM and Physics Department, University "La Sapienza", Rome, ITALY, 3INFM and Physics Department, University of L'Aquila, L'Aquila, ITALY, 4Department of Experimental Medicine and Pathology, University "La Sapienza", Rome, ITALY, 5Department of Cellular Biotechnology and Haematology, University "La Sapienza", Rome, ITALY. Introduction: Combined 1H and 13C NMR spectroscopy (MRS) provides powerful methods to elucidate biochemical pathways and monitor metabolic fluxes in biological systems. In this work we used a) 1H MRS to investigate the biochemical nature of mobile lipids (ML) formation in a lymphoblastoid cell line after treatment with the complex III inhibitor of mitochondrial electron transport antimycin A (amc-A) and b) a J-editing approach for the indirect detection of 13C nuclei, to monitor alterations of 13C label fluxes from [1-13C]-glucose to glycolytic intermediates and to ML formed after amc-A induced mitochondrial impairment. Methods: HuT 78 cells were exposed to amc-A (5 ug/ml) either in complete medium or in medium containing 5mM [1-13C]-glucose. Exposure to amc-A (24 h) did not alter cell viability nor induced apoptosis (<5%). MRS experiments were performed on intact cells and their extracts (organic and aqueous phases) on a Bruker Avance 400 spectrometer using a 1H-X multinuclear inverse probehead. Samples were analysed by both 1H MRS and a double resonance [13C]-1H technique which allows the detection of the only protons J-coupled to 13C nuclei by a sequence called T-Sedor [Casieri et al. Chem Phys Lett 338: 137-141, 2001]. This technique combines good sensitivity and chemical resolution which allow short

S280

Cells, Extracts, Fluids

measurement time to monitor the 13C-glucose metabolism in vivo. Lipid bodies were detected by fluorescence confocal laser scanning microscopy (CLSM) in cells stained with lipophilic dyes. Lipid analyses were performed by thin layer cromatography (TLC) on lipid extracts. Results and Discussion: After 24h amc-A treatment, 1H spectra of intact cells showed an over 10-fold increase of ML (CH2)n signal at 1.29 ppm and a well visible CH=CH signal at 5.35 ppm. TLC and CLSM analyses confirmed a strong accumulation of neutral lipids assembled in intracellular lipid bodies. T-Sedor spectra of amc-A-treated cells showed incorporation of 13C label from [1-13C]-glucose into (CH2)n and CH3 of ML signals, a 3-fold increase of the [3-13C]-lactate resonance and [13C]1H signals in the region between the 3.5 and 4.0 ppm. Analyses of cells extracts allowed identification in the 3.5-4.0 ppm region of glycolytic metabolites, such as glycero-3-phosphate and of phospholipids and beta-lyso-phospholipids (C-alpha) glycerol moiety. The [4-13C]-glutamate signal was present only in control cells where the mitochondrial oxidative capacity was not compromised. In conclusion, the capability of T-Sedor to select 1H-13C bonds from different molecules offers novel tools to monitor 13C fluxes from glycolysis to ML formation pathways in intact cells.

534 NMR analysis of hydration parameters of turbot ova, relationship with biological performance in aquaculture D. L. M. Pouliquen1, C. Munoz2, M. H. Omnes2, G. Tanguy3, J. L. Gaignon2; 1Inserm UMR 419, Institut de Biologie, Nantes, FRANCE, 2Lab. Physiologie des Poissons, Ifremer, Brest, FRANCE, 3Inserm ERM 0104, IBT, Angers, FRANCE. Purpose/Introduction: Fertilization and rearing of turbot (Psetta maxima) strongly depend on egg quality, which exhibits considerable variability. Attempts to identify the origin of this variability have led to numerous analyses of morphological and biochemical parameters (1), whose characteristics cannot be used to predict the egg's development well. The purpose of this study was to use the potential of temperature dependent NMR (2) to identify possible changes in the physical properties of bulk water related to egg's state of ripeness. Subjects and Methods: Twenty four batches of ova were collected from 6 different females from turbot broodstocks in Ifremer facilities. Sub-samples of ova were fertilized and incubated during 6 days at 13 °C to evaluate the hatching percentage of eggs. Determination of dry and wet weights, total lipids and proteins of the 24 batches of ova were performed in parallel to NMR measurements. Spin-lattice and spin-spin relaxation times of bulk water in ova (T1obs and T2obs) were measured at 5 decreasing temperatures from 23 °C to 4 °C, on a pc 140 minispec Bruker spectrometer. Results: The 24 batches of ova exhibited marked differences in their developmental pattern. Four main categories (UD, GD, GD+ and GD++) were distinguished with respect to increasing hatching rates (1.2, 46.8, 65.2, and 69.4, respectively) of the embryos. From UD to GD++, the temperature dependence of the spin-spin relaxation rate changed markedly, with comparable linear dependence of Log T2obs on 1/T but decreasing slopes corresponding to an activation energy Ed for the translational motion of water molecules of 1.95, 1.51, 1.45 and 1.04 Kcal. mol-1, respectively. A similar but less significant phenomenon was observed for T1obs. Simultaneously, a decrease in the protein to lipid ratio (4.91 to

4.54) and in the wet weight (0.76 to 0.67 mg / ovum) was observed from UD to GD++ groups. Discussion/Conclusion: The observed variations in the temperature dependence of the T2obs are consistent with modifications of the density of the macromolecular assemblies and organelles composing the yolk. This study has demonstrated that combination of basic biochemical analyses with temperature-dependent NMR measurements allows to distinguish batches of eggs with the highest quality of development. Acknowledgements: Part of this study was supported by Contract N° 97522181 from the Ifremer, Centre de Brest. References: 1. Scott, M. et al. World Aquaculture (1992) 23, 56. 2. Pouliquen, D. et al. Comp. Biochem. Physiol. (1998) 120B, 715.

Author Index A Abramovitch, R.: 250 Acebes, J. J.: 356 Adame, I. M.: 166 Adamek, D.: 106 Adami, A.: 155 Adriany, G.: 39 Aguilera, C.: 356 Ahearn, T. S.: 487 Ahlhelm, F.: 203, 84 Ajaj, W.: 145 Akhmetov, Y. A.: 496 Akkerman, E. M.: 187 Alahuhta, S.: 33 Alecci, M.: 422, 425 Alfke, H.: 162 Alfonsetti, M.: 425 Alfonsi, J.: 170 Ali-Cherif, A.: 63 Aliev, M. A.: 496 Alper, H.: 516, 517 Altay, H.: 352, 359, 516 Altes, T. A.: 169 Althaus, M.: 210, 394 Alvarez-Linera, J.: 467 Amann, M.: 186 Andersson, B.: 337, 340 Andersson, M.: 278 Anisimov, N. V.: 442 Annet, L.: 238, 492 Anthonsen, H. W.: 58 Anwander, A.: 454 Arana, E.: 372, 507 Arcuri, P. P.: 217, 299 Arcuri, V. M.: 217, 299 Armenean, M.: 424 Armitage, P. A.: 322, 528 Arnold, J. F. T.: 171 Aronson, L.: 195 Aroutunov, N. V.: 474 Arslanoglu, A.: 195, 251, 369 Arús, C.: 18, 314, 355, 356 Asahi, M.: 406 Aschwanden, M.: 233, 235 Aspelin, P.: 69 Atalar, E.: 461 Athanasiou, E.: 293 Atkinson, D.: 189 Audoin, B.: 63 Aumann, S.: 236 Avellino, A. M.: 195 Aytur, O.: 461

B Backes, W. H.: 231, 297, 57 Bacovský, J.: 510 Bader, D.: 308 Badiella, L.: 355 Baeli, I.: 83, 88 Baert, A.: 25 Bajbouj, M.: 198, 200 Balakrishnan, A.: 452 Ballesteros, P.: 267, 285, 416 Baltes, C.: 275 Balthazart, J.: 263 Banasik, T.: 106, 193, 287, 531

Banerji, U.: 403 Bankamp, A.: 457 Barbazeni, G.: 484 Barbiroli, B.: 397 Barceló, J.: 123, 300, 519 Barentsz, J.: 12 Barker, P. B.: 195, 251, 369 Barmet, C.: 125 Barral, F.: 353 Barreau, X.: 139 Barth, M.: 247, 261, 30, 326, 327, 36, 381, 5 Bartl, R.: 242 Bastin, M. E.: 322, 528 Batchelor, P.: 189 Battistoni, A.: 484 Baudendistel, K. T.: 168, 384 Bauer, W. R.: 205, 207, 208 Bauermann, T.: 427, 446 Bautista, D.: 507 Bax, J. J.: 202 Beavis, A. W.: 229 Beer, M.: 343, 42, 9 Beets, G. L.: 227 Beets-Tan, R. G. H.: 227, 297, 57 Behr, V. C.: 157, 164, 165, 410, 75 Beil, M.: 112 Belkic, D.: 365, 396 Bellemann, M.: 460 Bellot, G.: 409 Bells, S.: 31 Beloeil, J. C.: 286, 335, 77 Beloueche-Babari, M.: 131, 403 Benattayallah, A.: 162, 421 Bendahan, D.: 115, 524 Benderbous, S.: 453 Benito, M.: 267, 285, 416 Benoit, J. P.: 409 Bentas, W.: 228 Berardi, P.: 467 Berenschot, G.: 280 Berg, A.: 214, 82 Bergamini, C.: 467 Bernard, M.: 239, 81 Bernard, V.: 272 Berry, I.: 102, 94 Bertoldi, D.: 304 Bertoni, M. A.: 222 Beuf, O.: 424 Beusen, J.: 441 Bielen, D.: 23, 279 Bielicki, G.: 363 Bieri, O.: 434 Biglioli, F.: 234 Bilecen, D.: 233, 235 Bilgili, Y.: 376 Biltjes, I.: 292, 67 Bimson, W.: 64 Binder, M.: 387 Birsasteanu, F. F.: 521 Birtchnell, R.: 458 Bjerkvig, R.: 408 Blaimer, M.: 204, 40, 44, 445 Blaser, S.: 255, 46 Blezer, E.: 73 Bochynska, A.: 351 Bock, M.: 439, 457 Boehm, H. F.: 15

S281 Boehm, M.: 203 Boesch, C.: 110, 330, 332, 447, 47, 52 Boese, J. M.: 236 Boesiger, P.: 120, 125, 206, 275, 39, 391, 392, 41, 87, 99 Bogdahn, U.: 410 Boguslawska, R.: 351, 377 Bohic, S.: 415 Boichot, C.: 45 Bone, B.: 69 Bongartz, G.: 228, 233, 235, 306 Bongers, A.: 159 Bonifacio, C.: 234, 281 Bonny, J.: 107, 79 Bonomo, L.: 325 Börnert, P.: 276, 438 Borra, R. J. H.: 499 Borsato, A.: 501 Bosmans, H. T. C.: 279, 439, 443, 93 Botteman, F.: 412 Boujraf, S.: 147, 389 Boumans, T.: 262 Bouten, C.: 308 Boutry, S.: 56, 85 Bouzier-Sore, A. K.: 127 Bowtell, R. W.: 421 Bozzao, L.: 138 Braendli, M.: 306 Brahami, Y.: 453 Brami, C. T.: 349 Brandt, M. E.: 255 Branea, I.: 521 Brassen, S.: 134, 380 Braus, D. F.: 134, 380 Brekke, C.: 408 Brekken, C.: 408 Breuer, F.: 204, 40, 44, 445 Brockstedt, S.: 103, 188, 318 Brookeman, J. R.: 169 Broome, J.: 358 Brown, G.: 51 Bruggeman, F. L.: 116 Brummer, M. E.: 119, 383 Brunotte, F.: 45 Brurok, H.: 58 Bruvold, M.: 58 Brzegowy, P.: 106, 531 Buijs, J.: 280, 323, 379 Bulte, D. P.: 170, 431 Bureau, M. F.: 307 Burger, P. C.: 195 Burian, M.: 340 Burtea, C.: 133 Bus, S. A.: 144 Butterworth, R.: 126 Bystrický, P.: 201

C Cabañas, M.: 355, 356 Caccia, P.: 505 Cairns, B. E.: 143 Calamante, F.: 189 Calderan, L.: 155 Calhoun, V. D.: 251 Calli, C.: 352, 359, 511, 512 Candiota, A.: 355, 356 Canet Soulas, E. P.: 240, 454 Canioni, P.: 127, 59

Author Index

S282 Cantisani, R.: 495 Cappa, S.: 62 Capsa, R.: 295 Carbone, I.: 480 Carle, O.: 407 Carlier, P. G.: 290, 304, 307 Carme, S.: 240 Carnevale, R.: 533 Carota, A.: 66 Carpenter, T. K.: 346, 528 Caselli, S.: 480 Casieri, C.: 533 Cassinis, M.: 488 Cassol, E.: 102, 188 Castells, F.: 383 Catalaa, I.: 94 Catalano, C.: 477, 480, 485, 486 Caulo, M.: 325 Cavallo, T.: 495 Cavassila, S.: 328 Caverni, L.: 234 Cawkwell, L.: 71 Celda, B.: 361, 372 Celestre, M.: 418, 426, 500, 502, 83, 86, 88, 90 Cerdan, S.: 128, 267, 285, 416 Ceric, I.: 303 Chaabane, L.: 454 Chabiniok, R.: 220 Chahwan, M.: 351 Chassain, C.: 363 Chaturvedi, A.: 71 Chekenya, M.: 408 Chevrier, A.: 35 Chileg, S.: 243 Chinot, O.: 197 Chinzei, K.: 452 Choi, B.: 388 Chopier, J.: 483 Chung, G.: 386 Chung, S.: 388, 465 Chung, S. T.: 254 Cichon, A.: 362 Cicres, J.: 123 Cihak, R.: 230 Cinar, C.: 516 Ciosci, R.: 281 Cipolla, P.: 138, 493 Cirak, B.: 195 Clanet, M.: 102 Clarkson, R. B.: 414 Claus, F. G.: 191, 23 Claussen, C. D.: 92 Clementi, V.: 397 Coenradie, Y.: 328 Cognard, C.: 94 Cole, P.: 358 Collins, D. J.: 294, 403 Colosimo, C.: 139, 325 Colpaert, C.: 292, 67 Coltel, N.: 266 Colville, J. A. C.: 226 Comunale, L.: 501 Confort-Gouny, S.: 197, 266, 321, 63 Connelly, A.: 189 Cornalba, G.: 234, 281 Correze, J. L.: 286 Cortijo Mérida, M.: 419

Cosnard, G.: 324 Costa, S.: 507 Court, L.: 353 Cova, M.: 225 Cox, T.: 441 Cozzone, P. J.: 115, 197, 239, 266, 321, 524, 63, 81 Crémillieux, Y.: 489 Cucchiara, S.: 426 Cunnington, R.: 153, 257 Curt, A.: 154 Curto, C. A.: 422 Cvoro, V.: 322 Czapiga, E.: 378

D Dalagija, F.: 303 Danchaivijitr, C.: 219 Danchaivijitr, N.: 219 Daniel, G. H.: 70 Danielsen, E. R.: 282 Danilouchkine, M. G.: 202 Danti, M.: 477, 480, 486 Daprà, M.: 478 Darrasse, L.: 161, 334, 489 Das, K.: 358 De Bei, F.: 488 De Castro, S.: 480 De Catte, L.: 506 de Greiff, A.: 252, 432 de Groof, A.: 95 De Keyzer, F.: 191, 439 de Lange, E. E.: 169 De Luca, F.: 533 de Lussanet, Q. G.: 297, 57 De Reggi, M.: 266 De Ridder, D.: 514, 515 De Ridder, F.: 506 De Rochefort, L.: 489 de Roos, A.: 202 De Schepper, A. M.: 140, 141, 182, 273, 292, 43, 473, 514, 515, 518, 67 De Schutter, E.: 34 de Vries, H.: 73 De Wilde, J. P.: 464, 466 De Zeeuw, C. C. I.: 259 Deacon, R.: 112 Décorps, M.: 249 Dehoux, J.: 238 Deichmann, R.: 433 Delakis, I.: 466 Della Sala, S.: 62 Delorme, S.: 242 DelProposto, Z.: 244 Demirakca, T.: 370 den Heeten, G. J.: 284 Denis, B.: 321 Dennis, M.: 255, 322 Derbyshire, J. A.: 215 Derksen, S.: 379 Desrois, M.: 239, 81 Destro, G.: 481 Dezortova, M.: 135, 368, 449, 497 di Benedetto, P.: 501 Di Giorgio, D.: 325 Di Iorio, A.: 325 Di Roma, M.: 225 Didier, D.: 209

Didinger, B.: 186 Diergaten, T.: 92 Dietl, J.: 207 Dietrich, D.: 52 Dietrich, O.: 291 Dijkhuizen, R. M.: 406 Dijkstra, C.: 73 Dimicoli, J. L.: 243 Dimitriadou, E.: 327, 36, 381 Dimitrijevic, M. R.: 150 Dirnhofer, R.: 52 Dixon, W. T.: 119 Doan, B.: 335, 77 Dobhal, M. P.: 270 Dolenko, B.: 129 Dombrovsky, V. J.: 224, 503, 504 Donnat, J.: 107, 363, 79 Doornbos, J.: 202 Döpp, E.: 73 Dormont, D.: 286 Dornier, C.: 209 Dorsch, N.: 51 Douek, P.: 232, 454 Dreher, W.: 210, 312, 394 Drewes, A.: 298 Driesel, W.: 163 Drolsum, A.: 302 Drost, M. R.: 108 Drugova, B.: 520 Ducassou, D.: 272, 274 Duench, G.: 170 Duprez, T. P.: 324 Dupuich, D.: 489 Durand, C.: 45 Durand, E.: 489 Durif, F.: 363 Dvorak, P.: 337 Dydak, U.: 391, 41 Dymarkowski, S.: 177, 23 Dzeja, P.: 95 Dzik-Jurasz, A. S. K.: 403

E Edwards, A. D.: 341 Eeleta, M.: 373 Eerens, I.: 292, 67 Egan, G. M.: 72 Eggers, H.: 276 Eibel, H.: 283 Eidt, S.: 105 Elias, D.: 373 Elsner, K.: 460 Elster, C.: 400 Ende, G.: 136, 367, 370 Engelbrecht, M.: 12 Enger, P. O.: 408 Englund, E.: 318 Erb, M.: 331 Erginoz, E.: 517 Escanye, J.: 482 Essig, M.: 186, 374, 382 Eun, S.: 354 Eustace, S.: 14 Evangelista, P.: 218, 221

F Faber, C.: 157, 164, 165, 179, 336, 338, 410

Author Index Fabri, M.: 65 Fahlbusch, R.: 194, 196 Fantazzini, P.: 155 Farion, R.: 249, 268, 407 Farrall, A. J.: 322 Fautz, H.: 121 Favagrossa, G. M.: 505 Favilli, B.: 455 Felber, S. R.: 150, 151, 152 Felblinger, J.: 482 Fendrych, P.: 230, 368, 449, 525 Ferenci, P.: 498 Ferrante, A.: 404 Ferrari, R.: 83, 90 Ferri, M.: 90 Fezoulidis, I.: 293 Fidler, F.: 342 Fiegele, T.: 152 Figarella-Branger, D.: 197, 524 Fiori, L.: 281 Flach, H.: 319 Floris, S.: 73 Forjaz Secca, M.: 218, 221 Fornes, P.: 334 Forstner, R.: 301 Foxall, D.: 209, 476 Fraioli, F.: 477, 485 Franco, F.: 505 Francone, M.: 480 Franconi, J.: 59 Fransson, P.: 260 Frattini, T.: 493 Freitag, A.: 207 Frey, R.: 199 Fried, M.: 87 Fries, P.: 203, 84 Frøkjær, J. B.: 298 Fromes, Y.: 290, 304 Furlan, F.: 481 Fusenig, N. E.: 80 Futterer, J.: 12

G Gaignon, J. L.: 534 Gaily, E.: 402 Galanaud, D.: 197 Galant, J.: 180 Galbán, C. J.: 109, 145, 432 Gallinat, J.: 200 Gallucci, M.: 138 Galun, E.: 250 Galy-Lacour, C.: 274 Gambarota, G.: 143 Gandini, G.: 488 Ganslandt, O.: 194, 196 Gao, J.: 112 Garavaglia, C.: 155 Garcia-Alvarez, R.: 148 Gareis, D.: 157, 164, 165 Garlicki, A.: 371 Garmish, A.: 508 Garnett, M. R.: 3 Garnier, P.: 353 Garrido, S.: 128, 285 Garrido Salmón, C.: 419 Gartner, W. J.: 324 Gasparini, G.: 481 Gass, A.: 32

Gastaldelli, A.: 248 Gaviria, M.: 107, 79 Gavrankapetanovic, I.: 303 Gawehn, J.: 427, 446, 49 Gazeau, F.: 335 Geitung, J.: 302 Geoghegan, T. J.: 226 George, A. J. T.: 341 Georgescu, S.: 295 Georgi, J. C.: 462 Geppert, C.: 210, 394 Geraldes, C.: 267 Gérardin, E.: 139 Gerstenbrand, F.: 150 Ghanem, N.: 121 Gharib, B.: 266 Giacomini, E.: 429 Giannesini, B.: 115 Gibbs, P.: 448, 68 Gielen, J. L. M. A.: 140, 141, 24 Giesel, F. L.: 382 Gilbert, F. J.: 487 Gilderdale, D. J.: 158 Gilles, B.: 246 Gillespie, P.: 142 Gillet, B.: 286, 335 Gillies, R. J.: 128 Gilligan, P.: 72 Ginefri, J. C.: 161, 334 Girman, P.: 340 Gizewski, E. R.: 109 Gjesdal, K. I.: 302, 91 Gjikolli, B.: 303 Glaser, C.: 13 Glogarova, K.: 337 Glüer, C.: 55 Goisis, M.: 234 Golaszewski, S.: 150, 151, 152 Golebiowski, M.: 50 Gombas, W.: 261 Gomes, L.: 51 Gonçalves Pereira, P. M.: 218, 221 Gorbounov, N.: 301 Goris, R.: 523 Gortenuti, G.: 481, 484, 501 Gossuin, Y.: 451 Goto, T.: 428 Gounarides, J.: 112, 522 Gouyon, J.: 45 Goyen, M.: 232 Graesslin, I.: 276 Graf, H.: 237, 423 Gralla, J.: 435 Grandin, C. B.: 324 Granström, M.: 402 Grasso, G.: 501 Graveron-Demilly, D.: 328 Gray, H.: 128 Greenman, J.: 71 Greiser, A.: 338 Grenier, N.: 459 Griffioen, A. W.: 297, 57 Griffiths, J. R.: 305, 430, 436 Griswold, M. A.: 156, 19, 204, 37, 38, 40, 44, 445 Gröhn, O. H. J.: 104, 405 Gros, E.: 250 Gruber, S.: 194, 196, 199, 390

S283 Gubskii, L. V.: 350, 442 Guendisch, G. M.: 150, 151, 152 Guezenoc, F.: 139 Guihard, G.: 409 Guillot, G.: 489 Guis, S.: 524 Guney, B.: 516 Guzman, R.: 435 Gvozdjáková, A.: 201

H Haacke, M. E.: 244 Haas-Rioth, M.: 339 Haase, A.: 156, 157, 164, 165, 336, 338, 339, 342, 410, 469, 74 Hadac, J.: 368 Haddad, D.: 164, 339 Haddar, D.: 244 Hadzihasanovic, B.: 303 Hahn, D.: 343, 345, 42 Haigis, W.: 469 Haimerl, J.: 205 Hájek, M.: 135, 337, 340, 368, 401, 449, 497, 525 Hajnal, J. V.: 341, 450 Häkkinen, A.: 402 Hakumäki, J. M.: 269 Hallberg, L.: 48 Hamaekers, P.: 279 Hamilton, S.: 226 Han, Y.: 386 Hand, P.: 322 Hanson, L. G.: 137 Haraldseth, O.: 408 Harantová, P.: 135, 449 Harel, H.: 250 Haroon, H. A.: 348 Hartel, M.: 193, 287 Hartung, H.: 167 Haskard, D. O.: 341 Hasse, A.: 75 Haton, H.: 107, 79 Hayward, C.: 64 Heathcote, A.: 229 Heemskerk, A. M.: 108 Heerschap, A.: 114, 117, 12, 143, 211, 95 Hegde, S. R.: 458 Heidecker, H. G.: 233, 235 Heidemann, R. M.: 204, 37, 38, 40, 44, 445 Heijman, E.: 472 Heikkinen, S.: 402 Heiland, S.: 462 Heilmann, M.: 242, 80 Heinz, A.: 367 Helbing, W. A.: 60 Heller, M.: 55 Hempel, A.: 382 Hempel, E.: 382 Henn, F. A.: 136, 370 Hennenlotter, J.: 92 Hennerici, M. G.: 32 Hennig, J.: 118, 121, 216, 223, 347, 417, 444, 460 Henry, H.: 168 Hermans, R.: 191 Hermoye, L.: 238, 492 Herold, V.: 336

Author Index

S284 Herynek, V.: 337 Herzka, D. A.: 215 Hetzel, A.: 347 Heverhagen, J. T.: 162, 168, 384 Hiba, B.: 268 Hill, D. L. G.: 189, 458 Himmelreich, U.: 129, 51 Hirsch, J. G.: 32 Hoehn, M.: 317 Hofer, H.: 498 Hoffmann, G. F.: 47 Hoffmann, N.: 344 Homel, M.: 353 Hong, I. K.: 254 Horacek, J.: 61 Horák, D.: 510 Horecký, J.: 201 Horky, J.: 76 Hornik, K.: 327, 36, 381 Horska, A.: 195, 251, 369 Horsmans, Y.: 492 Horstick, G.: 344 Huber, S.: 469 Hubrich-Ungureanu, P.: 136 Huesler, J.: 52 Hurley, G.: 226 Hussain, S. M.: 100 Hutchison, J. M. S.: 468 Hyacinthe, J.: 476

I Iafrate, F.: 500, 86, 88, 90 Ibarrola, D.: 102, 239, 266, 63 Ilic-Jankovic, D. S.: 513 Ilic-Jankovic, D. S. T. V.: 513 Iltis, I.: 239, 81 in t Zandt, R.: 95 Indovina, P.: 404 Iorio, E.: 533 Iotti, S.: 397 Isbrandt, D.: 117 Isgaard, J.: 411 Issels, R.: 283 Ith, M.: 332, 52 Itzchak, Y.: 349 Ivancevic, M. K.: 209, 440, 476 Izquierdo, M.: 115, 239

J Jackson, A.: 28, 348 Jacobs, M. A.: 251 Jacobs, M. J. H. M.: 231 Jacquemyn, Y.: 506 Jaeger, K.: 233, 235 Jaillard, A.: 353 Jakob, P. M.: 156, 171, 204, 208, 342, 345, 37, 38, 40, 44, 445, 469, 53 Jamar, F.: 238 Jander, S.: 167 Janssen, E.: 95 Janssen, M.: 323 Jäntti, V.: 33 Januel, A.: 353 Jasinski, A.: 106, 193, 287, 531 Jendelova, P.: 337 Jeneson, J. A. L.: 116 Jeong, G.: 354 Jeong, S.: 386

Jeppsson, A.: 532 Jerosch-Herold, M.: 183 Jia, G.: 168 Jirák, D.: 340 Jiru, F.: 401 Jochimsen, T. H.: 146 Jolesz, F. A.: 452 Jones, N.: 51 Jordão, C.: 218, 221 Joshi, S. S.: 270 Jouanneau, J.: 241 Juchem, C.: 264 Judson, I. R.: 403 Jung, B.: 332 Junge, S.: 420 Jurczyk, A.: 378 Jynge, P.: 58

K Kacher, D. F.: 452 Kadi, N.: 483 Kaffy, J.: 241 Kaiser, W. A.: 247 Kan, H. E.: 117 Kanavou, T.: 293 Kang, H.: 354 Kaplanski, G.: 524 Kappas, K.: 293 Karg, A.: 344 Karlsson, M.: 475 Karlsson-Parra, A.: 411 Kasparová, S.: 201 Kasper, S.: 199 Kaufmann, I.: 74 Kauppinen, R. A.: 104, 405 Kautzner, J.: 230 Keevil, S. F.: 458 Kelarestaghi, M.: 453 Kellman, P.: 215 Kemper, C.: 452 Kendall, E.: 105 Kentgens, A. P. M.: 211 Kerskens, C.: 258 Kersschot, E.: 67 Kettunen, M. I.: 104, 405 Khan, S.: 464 Kiefer, C.: 435, 529 Kiessling, F.: 242, 80 Killeen, R. P. M.: 226 Kilzer, M.: 277 Kim, H.: 354 Kim, J.: 227, 89 Kim, M.: 227, 89 Kim, N.: 89 Kim, W.: 89 Kim, Y.: 254, 388, 465 Kindermann, I.: 203 Kirchin, M.: 139, 374, 478, 493 Kitis, O.: 352, 359, 511, 512 Kiviniemi, V.: 33 Kivistö, S. M.: 213 Kjellström, C.: 411 Kleinert, R.: 324 Klener, J.: 220 Klomp, D. W. J.: 114, 117, 143, 211 Klose, K. J.: 162 Klose, U.: 2, 237, 331 Kluger, W.: 214

Kneeshaw, P.: 68 Knopp, M. V.: 139, 168, 374, 384 Knutsson, H.: 278 Knysz, B.: 378 Kobayashi, A.: 351 Kober, F.: 239, 81 Kocaman, A.: 352 Koh, D.: 294 Kollias, S. S.: 147, 154,192, 389 Komarek, V.: 368 Komu, M.: 499 Konopka, M.: 193, 287 Kooi, M. E.: 111 Koopman, R.: 111 Koppelstaetter, F.: 150, 151, 152 Korach-Andre, M.: 112, 522 Köstler, H.: 343, 42 Kozerke, S.: 120, 206, 275, 392 Koziarski, A.: 360 Kozub, J.: 149, 371, 387 Kramann, B.: 203, 84 Krane, J.: 58 Krawczyk, R.: 351, 377 Kreis, R.: 110, 330, 332, 447, 47, 52 Kreitner, K. F.: 344 Krestin, G. P.: 132 Kristoffersen Wiberg, M.: 69 Kriz, J.: 340 Kroupova, J.: 337 Krsek, P.: 220, 368 Krssak, M.: 212, 498, 523, 526 Kucharská, J.: 201 Kuehn, B.: 291 Kuhl, C.: 21, 98 Kuliszkiewicz- Janus, M.: 378 Kuniecki, M.: 149 Kupsky, W.: 244 Kurc, T.: 384 Kushchaev, S.: 508 Kushnir, T.: 349 Kutschke, G.: 49 Kvitting, J. P. E.: 278 Kwiecinski, S.: 192 Kyriakos, W. E.: 479

L La Ferla, R.: 478 Ladd, M. E.: 109, 145, 252, 432 Laffon, E.: 272, 274 Laghi, A.: 11, 418, 426, 500, 502, 83, 86, 88, 90 Lahrech, H.: 407 Lalande, A.: 45 Lamalle, L.: 249 Lamb, H. J.: 202 Lambrecht, V.: 494 Lamerichs, R.: 41 Lamorgese, B.: 495 Lan, C.: 81 Lanciotti, K.: 480 Landini, L.: 248, 455 Lanfermann, H.: 320, 398, 399 Lange, O.: 276 Lange, T.: 391 Langereis, S.: 57 Laniado, M.: 310 Laporte, S.: 353 Larkman, D. J.: 341, 450

Author Index Larsen, E.: 298 Larsson, E. M.: 103, 188, 29, 260, 309, 318 Lätt, J.: 103, 188, 318 Lauerma, K.: 213 Laumonier, C.: 85 Laurent, D.: 112, 522 Laurent, F.: 272, 274 Laurent, S.: 133, 412, 413, 415 Lautamäki, R.: 499 Lavini, C.: 284 Lazeyras, F.: 476, 66 Le Bihan, D.: 17, 429 Le Fur, Y.: 197, 266, 321, 524 Le Roux, P.: 482 Leach, M. O.: 131, 294, 403, 466 Ledru, F.: 334 Lee, S.: 386 Leemans, A.: 530 Leibfritz, D.: 126, 210, 394 Leiner, T.: 231 Leixnering, M.: 523 Lelieveldt, B. P. F.: 166 Lenti, L.: 533 Leroy-Willig, A.: 304, 307 Lethimonier, F.: 429 Leupold, J.: 434 Levi, M.: 144 Levine, A. L.: 168 Levine, B.: 255 Li, K. L.: 348 Li, M. H.: 70 Li, W. B.: 70 Liimatainen, T. J.: 104, 269 Lim, J.: 89 Lind, M.: 71 Lindahl, A.: 532 Lindner, L.: 283 Liney, G. P.: 142, 148, 229 Lingawi, S. S. A.: 375, 509 Link, A.: 400 Link, T. M.: 15 Lipp, M.: 343 Liptaj, T.: 201 Llach, S.: 123 Lo, E. H.: 406 Lococo, E.: 533 Lodemann, K.: 139 Lodi, R.: 397 Loevblad, K. O.: 529 Logie, R.: 62 Logothetis, N. K.: 264 Lohninger, A.: 498 Lomas, D. J.: 20 Lombardi, M.: 248, 455 Lopez, A.: 453 López, P.: 416 Lorentzon, M.: 532 Losert, C.: 291 Lowe, M. J.: 32 Lowry, M.: 448, 71 Loytved, W.: 84 Ludwig, C.: 436 Ludwig, U. A.: 121 Lukes, D. J.: 411 Lundbom, J.: 402 Lundbom, N.: 402 Lundborg, G.: 260

Lundervold, A.: 408 Lundgren, A.: 411 Lundstrøm, K.: 282 Lupescu, I.: 295 Lutz, N. W.: 130 Lutz, T.: 47 Luypaert, R.: 245 Lysiak, Z.: 351 Lyszcz, P.: 7

M Maas, M.: 144 Maass, J.: 488 Macagno, M.: 488 Mach, T.: 371 Machann, J.: 113 Mächler, H.: 7 MacPherson, S.: 62 Mader, I.: 331 Maderwald, S.: 109, 145, 432 Maes, F.: 279 Magenta Biasina, A.: 234, 281 Magistretti, P.: 127 Magnenat-Thalmann, N.: 246 Magnin, I. E.: 454 Maier, S. E.: 192 Majcher, K.: 106, 531 Mäkiranta, M. J.: 33 Malikova, I.: 63 Mamata, H.: 192 Mandiyan, S.: 522 Mandry, D.: 482 Manelfe, C.: 102, 94 Manenti, G.: 225 Manjon-Herrera, J.: 463 Mann, K.: 367 Mannfolk, P.: 29 Manning, W. J.: 125 Manor, D.: 349 Mansard Desbleds, C.: 454 Mansour, M.: 325 Manton, D. J.: 68, 71 Manzoni, T.: 65 Maraveyas, A.: 71 Marchal, G.: 23, 279, 439, 93 Marchi, A.: 484 Marie, P.: 482 Marin, D.: 418, 426, 500, 83, 86, 88, 90 Marsault, C.: 483 Martegani, A.: 493 Martel, A. L.: 456 Marthan, R.: 272, 274 Martí-Bonmatí, L.: 361, 372, 383, 463, 490, 507 Martín, A. B.: 267 Martinez, M. J.: 490 Martinez de Vega, V.: 176 Martinez-Bisbal, M.: 361, 372 Martinovic, J.: 219 Martirosian, P.: 237, 423, 92 Marusic, P.: 368 Marzola, P.: 155 Mateiescu, S.: 145, 432 Mathiesen, H. K.: 137 Matot, I.: 250 Matschl, V.: 162 Mattei, J.: 524 Mattle, H.: 529

S285 Matulewicz, L.: 362 Mayes, A.: 219 Mayo, K. H.: 297 Mazov, N. V.: 350 Mazza, G. C.: 505 McAuliffe, M. J.: 384 McDonald, M. A.: 414 McGregor, O.: 170 McIntyre, D. J. O.: 305, 430, 436 McVeigh, E. R.: 215 Meca, C. C.: 429 Medina, E.: 221 Meflah, K.: 409 Meier, D.: 41 Meigen, T.: 469 Meijer, E. W.: 57 Melberg, A.: 48 Memis, O.: 461 Mentlein, R.: 55 Meric, P.: 286, 335 Merkle, H.: 163, 264 Meuli, R.: 393 Meyer, J.: 344 Meyerspeer, M.: 212, 498, 526 Michalak, E.: 288 Michalak, M. J.: 288 Michel, D.: 353 Michel, S. C. A.: 313 Michels, R. P. J.: 144 Michiels, J.: 439 Miglio, C.: 418 Milano, F.: 471 Mildner, T.: 146 Miller, W.: 169 Millet-Roig, J.: 119, 383 Miquel, M. E.: 458 Miraux, S.: 59 Missert, J.: 270 Mitschang, L.: 395 Mitsouras, D.: 479 Mittelbach, M.: 276 Mlcochova, H.: 230 Mlynárik, V.: 199, 201, 390, 526 Moccozet, L.: 246 Mochizuki, H.: 428 Moedder, U.: 167 Moehler, T.: 242 Mohamed, M.: 166 Mollá, E.: 507 Monleon, D.: 361 Montaudon, M.: 272 Montemezzi, S.: 481, 484, 501 Montet, X.: 476 Moody, A. R.: 456 Moonen, C.: 172, 316, 459 Moore, A.: 46 Moore, L.: 294, 466 Moratal-Pérez, D.: 119, 383 Moratalla, R.: 267 Mörchel, P.: 336 Moreau, C.: 409 Morgan, P. S.: 456 Moser, E.: 153, 194, 196, 199, 212, 214, 247, 257, 261, 277, 30, 326, 327, 36, 381, 390, 498, 526, 82 Motoi, S.: 521 Motta, A.: 404 Mottaghy, F. M.: 150, 151, 152

Author Index

S286 Mougenot, C.: 459 Mountford, C. E.: 129, 51 Mueller, E.: 437 Mueller, M. F.: 156 Mugler, J. P.: 169 Mulder, W. J. M.: 472, 54 Mulkern, R. V.: 479 Müller, M.: 204, 40, 44, 445 Muller, R. N.: 133, 412, 413, 414, 415, 451, 56, 85 Munoz, C.: 534 Muñoz Maniega, S.: 322 Munter, K.: 441 Murray, A. D.: 385 Muthurangu, V.: 458 Mutignani, M.: 495 Mych, M.: 362 Myslivecek, M.: 510

N Nag, D.: 142 Nagai, S.: 428 Nägele, T.: 331 Nahrendorf, M.: 208, 338 Nardis, P.: 477 Navarro, D.: 126 Nazarpoor, M.: 456 Nedeltchev, K.: 529 Neeman, M.: 27 Nekula, J.: 510 Netsch, T.: 276 Neuberger, T.: 157, 164, 165, 338, 410 Neuburger, T.: 75 Neumann-Haefelin, C.: 320 Neumann-Haefelin, T.: 320, 398, 399 Neyran, B.: 240, 454 Nguyen, N. X.: 469 Nicolato, E.: 155 Nicolay, K.: 108, 111, 124, 308, 472, 54, 73 Nicoli, F.: 197, 321, 524 Nielles-Vallespin, S.: 457 Nielsen, A. H.: 296, 298 Nijenhuis, R. J.: 231 Nijs, E.: 23 Nikolaou, K.: 291 Nimsky, C.: 194, 196 Nittka, M.: 291 Nonaka, M.: 428 Nordhøy, W.: 58 Norris, D. G.: 146, 258 Nosàs, S.: 242 Noseworthy, M. D.: 170, 255, 31, 35, 431, 46 Nowé, V.: 514, 515 Nuutila, P.: 499

O O'Brien, M.: 305, 430 O'Donnell, S.: 255 Oakden, W.: 46 Obenaus, A.: 105 Occhiato, R.: 485, 486 Ochi, H.: 428 Odille, F.: 482 Oechsner, M.: 157, 164, 165 Oelschläger, H. H. A.: 339 Oesterle, C.: 460

Ogresta, F.: 373 Oh, C.: 388, 465 Oh, Y.: 89 Olausson, M.: 411 Oliete, S.: 467 Oliveira, E.: 218 Olsrud, J.: 260, 29 Olt, S.: 469, 53 Omerovic, E.: 532 Omnes, M. H.: 534 Oncel, D.: 289 Oncel, G.: 289 Onet, D. I.: 521 Oomens, C.: 308 Op de Beeck, B. J.: 10, 494, 506 Op de Beeck, K.: 23 Orkisz, M.: 454 Osteaux, M.: 245, 494 Ostergaard, L.: 406 Oswald, H.: 529 Ouahabi, A.: 453 Ouhlous, M.: 319 Owers-Bradely, J.: 421 Özsarlak, Ö.: 182, 22, 273, 43, 473, 518

P Padberg, G. W.: 114 Palussiere, J.: 459 Pandey, R. K.: 270 Panebianco, V.: 502 Pantano, P.: 138 Panzarella, M.: 455 Paolantonio, P.: 418, 426, 500, 86, 88, 90 Parizel, P. M.: 140, 141, 182, 273, 43, 473, 494, 506, 514, 515, 518 Parkes, L. M.: 258 Parkkola, R.: 499 Parzy, E.: 290 Pasan, A.: 436 Passariello, R.: 418, 426, 477, 480, 485, 486, 500, 502, 83, 86, 88, 90 Patankar, T. A.: 348 Patrascu, J.: 521 Pattynama, P. M. T.: 60 Paulson, O. B.: 137 Payen, J. F.: 407 Payne, G. S.: 403 Pedachenko, E.: 508 Pedersen, T. B.: 408 Pediconi, F.: 477, 485, 486 Peeters, F.: 238, 324, 492 Peeters, R.: 279 Pegios, W.: 228 Pelikánová, T.: 525 Peller, M.: 283, 291 Pellerin, L.: 127 Pelletier, J.: 524, 63 Pelliccia, A.: 480 Pellissier, J.: 524 Peñaloza, F.: 490 Penet, M.: 266 Pennell, D.: 174 Pérez de Alejo Fortún, R.: 419 Pérez-Alvarez, F.: 123 Perhonen, M. A.: 213 Perin, F.: 243 Perrin, R. L.: 246, 440 Perthel, M.: 7

Petersen, S. E.: 344 Petite, H.: 335 Petz, R.: 248 Pfeuffer, J.: 264 Pictet, J.: 393 Pieniazek, P.: 193, 287 Pietz, J.: 447, 47 Pilatus, U.: 320, 398, 399 Pilc, A.: 106 Pilkowska, E.: 377 Pingitore, A.: 455 Piquer, J.: 361 Pirogov, Y. A.: 442 Pirovano, G.: 374, 478 Place, F.: 463 Placidi, G.: 422, 425 Plessis, M. J.: 243 Podo, F.: 533 Podsiadlo, L.: 149, 371 Poenaru, D. V.: 521 Poirier-Quinot, M. L.: 161, 334 Polonara, G.: 65 Polzer, H.: 168 Pompili, G.: 234, 281 Poniatowska, R.: 351, 377 Porcaro, A.: 501 Porter, D. A.: 437 Portha, B.: 81 Portnoy, L.: 491 Positano, V.: 248, 455 Pottier, E.: 335 Pouliquen, D. L. M.: 409, 534 Poultsidou, A.: 293 Pozzi Mucelli, R.: 225 Pracht, E. D.: 171 Preiss, M.: 61 Privat, A.: 107, 79 Prokop, M.: 232 Provent, P.: 268 Pruessmann, K. P.: 147, 16, 206, 39 Pujol, J.: 123 Puppini, G.: 481 Purea, A.: 339, 75

Q Quick, H. H.: 109 Quintero, M. R.: 355, 356

R Ragab, Y.: 375, 509 Rainaldi, G.: 404 Raininko, R.: 48 Rakotonirainy, O.: 411, 532 Rakowicz, M.: 377 Ramaprasad, S.: 265, 270 Ramoni, C.: 533 Ramos Cabrer, P.: 124 Rand, T.: 82 Ranjeva, J.: 197, 321, 63 Ranjeva, J. P.: 102 Ratiney, H.: 328 Rating, D.: 47 Rauch, A.: 205, 207, 208, 345 Rauscher, A.: 247, 261, 30 Raybaud, C. A.: 97 Razavi, R.: 120, 458 Recheis, W.: 152 Redpath, T. W.: 487

Author Index Rees, M. R.: 185 Regan, F.: 226 Regnicolo, L.: 138 Reiber, J. H. C.: 166, 202 Reichenbach, J. R.: 244, 247 Reinfried, L.: 199 Reinl, H. M.: 283 Reiser, M. F.: 283, 291 Reiter, G.: 7 Reiter, U.: 7 Reitz, A.: 154 Remonda, L.: 435, 529 Rémy, C.: 249, 268 Renema, W. K. J.: 117, 95 Renou, J.: 107, 363, 79 Resmer, F.: 468 Reus, U.: 55 Reutelingsperger, C. P. M.: 315 Reutenauer, H.: 407 Reynolds, P. R.: 341 Rhomberg, P.: 151, 152 Richter, M. C.: 256 Richter, W.: 256 Rick, J.: 347 Rienmüller, R.: 184, 7 Riera, E.: 519 Ries, M.: 62 Rieta, J. J.: 383 Righetti, A.: 209 Rigotti, G.: 484 Rijpkema, M.: 114 Rinneberg, H.: 160, 198, 200, 395, 400 Risse, F.: 236 Ritter, C.: 343, 42 Rivers, C. S.: 322 Riviere, C.: 335 Roberts, C.: 471 Roberts, N.: 190, 219, 64 Robinson, S. D.: 261 Robles viejo, M.: 463 Roch, A.: 414, 451 Roden, M.: 498, 523, 526 Rodrigues, T. B.: 128, 267, 285, 416 Rodríguez, I.: 419 Rogozhyn, V.: 364, 366, 508 Rolencová, E.: 230, 449 Romano, R.: 404 Rommel, E.: 336 Ronen, S. M.: 131, 403 Roos, G.: 280, 323, 379 Roselli, A.: 486 Rosén, B.: 260 Rosen, B. R.: 406 Rosol, T. J.: 168 Rossmanith, C.: 32 Rouze, D.: 415 Rozhkova, Z.: 364, 366, 508 Rubió, A.: 519 Rudin, M.: 96 Ruf, M.: 134, 136, 380 Ruiz-Cabello, J.: 419 Runge, V. M.: 374 Ruohonen, J.: 33 Rybicki, F. J.: 479 Ryterski, J.: 351 Ryu, W. S.: 254 Ryu, Y.: 388, 465 Rzepka, E.: 270

S Sachs, G.: 261 Sailer, J.: 82 Saint-Jalmes, H.: 424 Saleh, A.: 167 Salerno, M.: 169 Salgado, R.: 506 Salomäki, T.: 33 Salomir, R.: 459 Saltz, J. H.: 384 Salvolini, U.: 65 San Roman, J.: 373 Sanchez, P.: 267 Sandstede, J.: 343, 42 Sansoni, I.: 500, 502, 83, 86, 90 Sansonnetti, P.: 77 Santarelli, M. F.: 248, 455 Santini, M.: 404 Saris, W. H. M.: 111 Sasamata, M.: 406 Sasiadek, M.: 378 Saudek, F.: 340 Sauer, D.: 205 Sauer, E.: 205, 207 Savas, R.: 516, 517 Scarano, E.: 495 Schachar, R.: 35 Schad, L. R.: 122, 159, 236, 457, 80 Schaeffter, T.: 276 Schanzenbächer, P.: 208 Schär, M.: 392 Scheenen, T. W. J.: 12, 143 Scheffler, K.: 434, 439 Schelfout, K.: 67 Scherman, D.: 307 Scheurer, E.: 52 Schick, F.: 113, 237, 423, 527, 92 Schima, W.: 181 Schlager, A.: 151 Schlemmer, H.: 92 Schmidt, M.: 339 Schmitt, A.: 134, 380 Schmitt, M.: 344 Schneider, F. C. G.: 353 Schneider, G.: 203, 232, 84 Schneider, P.: 283 Schoenberg, S. O.: 291 Schrauwen-Hinderling, V. B.: 111 Schreiber, W. G.: 344 Schroeder, C.: 438 Schroeder, J.: 382 Schroeter, M.: 167 Schroth, G.: 529 Schubert, F.: 198, 200, 395, 400 Schulte, R. F.: 391 Schultz, G.: 345 Schunter, J.: 200 Schurch, B.: 154 Schwenk, S.: 32 Schwizer, W.: 87 Scifo, P.: 429 Sciot, R.: 141 Sclavi, N. E.: 222 Seeger, U.: 331 Segebarth, C.: 6 Segers, J.: 85 Seghier, M. L.: 66 Seguin, J.: 307

S287 Sehgal, V.: 244 Seidel, R.: 84 Seidl, U.: 382 Seifert, F.: 160, 198, 200, 400 Semple, S. I. K.: 487 Sener, N. R.: 512 Seo, J.: 354 Serduc, R.: 268 Serikova, S. E.: 496 Serrao, F.: 217, 299 Seton, H. C.: 468 Shamalov, N. A.: 350 Shen, J.: 470, 78 Sherman, K. P.: 142 Shibaeva, A. N.: 350 Shimizu, Y.: 326 Siedentopf, C. M.: 150, 151, 152 Siegler, P.: 122 Sierra, A.: 267, 285, 416 Sigfridsson, A.: 278 Simmons, A.: 471 Simonetti, G.: 225 Singer, O.: 320 Sinitsyn, V.: 175, 8 Sirin, H.: 352 Sironi, A. M.: 248 Sjöbeck, M.: 318 Škoch, A.: 401, 449, 525 Skórka, T.: 106, 193 Skrobowska, E.: 357, 360 Skrzynski, S.: 357 Sloan, A.: 244 Slocum, A.: 452 Slotboom, J.: 529 Smans, K.: 279 Smith, E. T. S.: 358 Smits, J.: 57 Sobiecka, B.: 149, 371, 387 Söderman, O.: 318 Softic, M.: 303 Sokol, M.: 362 Sokolov, K. V.: 350 Sokolska, V.: 378 Sommers, M. G.: 178 Somorjai, R.: 129 Somsen, A. G.: 209 Sonkajärvi, E.: 33 Sorensen, A.: 406 Sørensen, P. S.: 137 Sorrell, T. C.: 129, 51 Sotgiu, A.: 422, 425 Sourbron, S.: 245 Soussi, B.: 411, 532 Spadaro, F.: 533 Spaniel, F.: 135 Speck, O.: 121, 223, 347 Sperl, J.: 497 Spira, G.: 250 Splendiani, A.: 138 Squillaci, E.: 225 Stadlbauer, A.: 194, 196, 199, 390 Ståhlberg, F.: 103, 471 Starcuk, Z.: 76 Starcuk, Z., Jr.: 390, 76 Starcukova, J.: 76 Starkel, P.: 492 Stashuk, G.: 491 Steffensen, E.: 298

Author Index

S288 Steidle, G.: 527 Steinbrich, W.: 228 Steingoetter, A.: 87 Stekelenburg, A.: 308 Stengel, A.: 320, 398, 399 Stenzl, A.: 92 Sterr, A. M.: 64 Stieltjes, B.: 186 Stoeter, P.: 427, 427, 446, 49 Storaas, T.: 91 Storas, T.: 302 Storm, G.: 54 Strijkers, G. J.: 108, 311, 472, 54 Strom, R.: 533 Stroman, P. W.: 193 Stubbe, I.: 245 Stuber, M.: 125 Stumvoll, M.: 113 Stupar, V.: 489 Sulek, Z.: 106 Sulek, A.: 377 Sumbalová, Z.: 201 Sumii, T.: 406 Summers, P. E.: 147, 192, 389 Sun, D.: 112 Sunaert, S.: 101, 191, 93 Suominen, K.: 33 Sykova, E.: 337 Szabo, B. K.: 69 Szeszkowski, W.: 50

T Tabakov, S.: 471 Taimr, P.: 497 Takahashi, T.: 428 Takeuchi, H.: 428 Takizawa, M.: 428 Tallheden, T.: 532 Talos, F.: 452 Tancredi, D.: 502 Tanguy, G.: 534 Taniguchi, Y.: 428 Tarnacka, B.: 50 Tartaro, A.: 138, 325 Taylor, A. M.: 458 Taylor, J.: 51 Teichert, N.: 283 Tellez, H. J.: 451 Ten Hove, M.: 271 ter Haar Romeny, B. M.: 472 Terrier, F.: 246 Tervonen, O.: 33 Terzic, A.: 95 Testa, C.: 533 Thacker, N. A.: 253, 329, 333 Thamer, C.: 113 Thelen, M.: 344 Thiaudière, E.: 59 Thoeny, H. C.: 191 Thomas, C. D.: 241, 243 Thomsen, C.: 282 Thunberg, P.: 475 Thurner, S.: 326 Tihan, T.: 195 Tindemans, I.: 263 Tintera, J.: 135, 220, 230, 427, 446, 61 Tivesten, Å.: 411 Tjalma, W.: 292

Tomanek, B.: 193 Tommasi, J.: 155 Tong, K.: 244 Topgaard, D.: 318 Torreggiani, W. C.: 226 Tost, H.: 134, 380 Tournebize, R.: 77 Tournier, D.: 189 Tozer, D.: 71 Trabesinger, A. H.: 391 Treier, R.: 87 Troprès, I.: 249 Tsai, J. C.: 75 Tsao, J.: 120, 206 Tscherning, T.: 137 Turnbull, L. W.: 148, 448, 68, 71 Turner, R.: 26, 4, 433 Tyler, K.: 358

U Uffmann, K.: 145, 432 Ugurbil, K.: 39 Umathum, R.: 457 Unal, B.: 376 Urbanik, A.: 149, 193, 371, 387 Urdzikova, L.: 337

V v. Mengershausen, M.: 146 Vallée, J.: 209, 246, 440, 476 Vallet, C.: 272 Vallette, F. M.: 409 Vallone, B.: 488 Valonen, P.: 104 Valverde, D.: 355 van Asten, J. J. A.: 114 Van Audekerke, J.: 262 Van Beers, B. E.: 238, 492 van Bloois, L.: 54 Van Breuseghem, I.: 443 Van Camp, N.: 259, 34 Van de Heyning, P.: 514, 515 Van de Moortele, P.: 39 van den Berg, J.: 60 van den Bos, E.: 132 van der Geest, R. J.: 166, 202 van der Graaf, M.: 114, 211 van der Klink, J. J.: 393 Van der Linden, A.: 259, 262, 263, 34 van der Lugt, A.: 319 van der Meulen, Y. M.: 211 van der Pol, S.: 73 van der Schaft, D. W. J.: 297 van der Toorn, A.: 124 van Dijk, L. C.: 319 van Dorsten, F.: 12 van Duynhoven, J. P. M.: 124 Van Dyck, P.: 273 Van Echteld, C. J. A.: 271 van Eggermond, J.: 438 van Engelshoven, J. M. A.: 227, 231 van Engelshoven, J. M. H.: 297, 57 van Genderen, M. H. P.: 57 van Geuns, R. J.: 60 Van Goethem, J. W.: 182, 43, 473, 518 Van Goethem, M. L. A.: 292, 67 van Loon, L. J. C.: 111 Van Meir, V.: 262

van Ormondt, D.: 328 van Pul, C.: 280, 323, 379 van Sambeek, M. R. H. M.: 319 van Weerelt, M.: 494 van Westen, D.: 260, 29 Vanbeckevoort, D.: 23 Vander Elst, L.: 133, 412, 413, 415, 443, 56, 85 Vanderdood, K.: 494 Vanhoenacker, F.: 140, 141 Vanzulli, A.: 232 Vaquero, J.: 419 Vassiou, K.: 293 Veenland, J. F.: 319 Vega, M.: 507 Venditti, F.: 477, 485, 486 Verhoye, M.: 259, 262, 263, 34 Verius, M.: 150 Vermathen, P.: 110, 330, 447 Verslegers, I.: 67 Verslegers, I. O. J.: 292 Versteegh, M. I. M.: 202 VIDAL, C.: 286 Vignaud, A.: 489 Vilanova, A.: 280 Vilanova, J. C.: 123, 300, 519 Vilanova i Bartroli, A.: 108, 472 Villa Valverde, P.: 419 Villalón, M.: 300, 519 Villarreal, E.: 490 Viola, A.: 266 Viout, P.: 197 Visconti, S.: 502 Vít, H.: 340 Vlak, M.: 114 Vliegen, R. F. A.: 227 Vlieger, E. J.: 187 Vlodavsky, I.: 250 Vlychou, M.: 293 Vogl, T.: 228 Vogt, F.: 252 Voisin, P.: 127 Volk, A.: 241, 243 Volny-Luraghi, A.: 34 Von Euw, D.: 290, 304 von Meyenfeldt, M. F.: 227 von Schulthess, G. K.: 173 von Tengg-Kobligk, H.: 168, 384 Vosseler, S.: 80 Votta, V.: 485 Vroemen, M.: 410 Vucurevic, G. V.: 49 Vuilleumier, P.: 66 Vuori, K.: 402 Vyhnanovská, P.: 525 Vymazal, J.: 220

W Waagstein, F.: 532 Wacker, C. M.: 208 Wagner, H. J.: 162 Waiter, G.: 62 Waiter, G. D.: 385 Walczak, C.: 241, 243 Waldman, A. D.: 1 Walecki, J.: 288 Walker, P. M.: 45 Walter, S.: 367, 370

Author Index Walzel, M.: 400 Wang, T.: 171, 208, 345 Wang, X. L.: 140, 141, 515 Warczynska, A.: 357, 360 Wardlaw, J. M.: 322, 346 Warfield, S. K.: 452 Wary, C.: 304 Wasser, K.: 242 Wasserman, B. A.: 166 Watkin, K. L.: 414 Webb, A.: 156, 338, 75 Weber, J.: 223 Weber, T.: 165, 410, 75 Weber-Fahr, W.: 136, 367, 370 Weglarz, W. P.: 106, 287, 531 Wegner, L.: 74 Weidner, N.: 410 Weigel, M.: 118, 121, 216, 223, 417, 444 Weis, J.: 48 Weishaupt, D.: 87 Welch, A. E.: 487 Welzel, H.: 367, 370 Wendt, C.S.: 134, 380 Weniger, D.: 389 Wernicke, C.: 200 Westenberg, J. J. M.: 202 Weyler, J.: 141, 67 White, M. J.: 294 Wieland, I.: 157, 164, 165, 410 Wielopolski, P. A.: 132, 60 Wieringa, B.: 95 Wieshmann, U.: 190, 219 Wiesinger, F.: 39 Wigström, L.: 278, 475 Wijn, P.: 280, 323, 379 Wildgruber, D.: 370 Wilhelm, T.: 186 Willemsen, M. A. A. P.: 211 Williams, J. H. G.: 385 Williams, S. R.: 253, 329, 333 Williamson, D. C.: 253, 329, 333 Wilmink, J. T.: 231 Wilms, G. E. R. C.: 93 Windischberger, C.: 153, 257, 261, 277, 30, 326 Wirestam, R.: 103, 471 Witoszynskyj, S.: 247 Wolf, I.: 134, 380 Wolter, J.: 55 Workman, P.: 403 Wu, D. H.: 70 Wu, O.: 406 Wübbeler, G.: 160 Wuestenberg, T.: 382 Wurm, T.: 92 Wuyts, F. L.: 514

Y Yahia-Cherif, L.: 246 Yilmaz, T.: 511, 512 Ylä-Herttuala, S.: 104, 269 Yousem, D. M.: 195 Yuh, W. T. C.: 70 Yunten, N.: 352, 359, 511, 512 Yurtseven, T.: 359, 511, 512

Z Zaharia, C.: 295

Zaman, A.: 190, 219, 64 Zamorano, L.: 244 Zanella, F.: 320, 398, 399 Zaremba, J.: 377 Zdzienicka, E.: 377 Zehnder, M.: 332 Zenga, A.: 281 Zhang, H.: 154 Zhang, Z.: 132 Zhu, X. P.: 348 Ziegler, A.: 268 Zielinski, G.: 360 Zimine, I.: 209, 476 Zimmermann, U.: 74 Zink, M.: 7 Ziyeh, S.: 223, 347 Zuna, I.: 242 Zwingmann, C.: 126 Zwygart, K.: 332, 47

S289